- Email: [email protected]
Mineral image enhancement based on sequential combination of toggle and top-hat based contrast operator
Micron 44 (2013) 193–201
Contents lists available at SciVerse ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
Mineral image enhancement based on sequential combination of toggle and top-hat based contrast operator Xiangzhi Bai ∗ Image Processing Centre, Beijing University of Aeronautics and Astronautics, 100191 Beijing, China
a r t i c l e
i n f o
Article history: Received 13 March 2012 Received in revised form 7 June 2012 Accepted 17 June 2012 Keywords: Toggle contrast operator Top-hat based contrast operator Mineral image enhancement Mathematical morphology
a b s t r a c t Enhancing mineral image especially making mineral image details clear is very useful for mineral analysis. To effectively enhance mineral image, an algorithm based on the toggle contrast operator and top-hat based contrast operator is proposed in this paper. Sequentially combining the toggle contrast operator and top-hat based contrast operator could be used to identify image features especially the image details. So, appropriately exacting the identified image features by the sequentially combined toggle and tophat based contrast operator is important for mineral image enhancement, which is analyzed firstly in this paper. After that, the multi-scale extension of feature extraction is given and used to construct the final features for mineral image enhancement. By importing the final extracted image features into the original mineral image through contrast enlargement, the original mineral image is well enhanced and the mineral image details are very clear. Experimental results on different types of mineral images verified the effective performance of the proposed algorithm. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Microscopy is the important instrument for mineral analysis (Kaiser and Chuvilin, 2003; Shah, 2007; Lumpkin et al., 1997). Obtained mineral images from different types of microscopes, such as scanning electron microscopy, scanning tunneling microscopy and transmission electron microscopy, could provide valuable information for mineral property analysis, new mineral identification, interactions analysis of different minerals and so on (Schultze-Lam et al., 1992; Carter et al., 2000; Jones et al., 2007). Designing microscopy is an effective way to obtain clear images (Schilders et al., 1998; Reshak, 2009). However, because the operators of microscopy equipment may not well produce the image of mineral samples or the imaging procedure is too difficult to produce clear mineral images, many obtained mineral images do not have good contrast and the contained details are usually not clear. These may affect the applications of these mineral images for the further mineral analysis. Therefore, it is useful to well enhance the mineral images and makes the image details clear. Image enhancement technique, which is an active research area in image processing, is the effective way for mineral image enhancement (Bai and Zhou, 2011; Heintzmann et al., 2003). Histogram based algorithms (Huang et al., 2006; Wan and Shi, 2007) are widely used and effective ways for image enhancement. But,
∗ Tel.: +86 1082338578. E-mail address: [email protected] 0968-4328/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.micron.2012.06.009
the bright image regions in mineral image may be over enhanced. Diffusion based algorithms (Tang et al., 2001; Gilboa et al., 2004) are mainly focusing on noise reduction, which is useful for noise smoothing. Filter based algorithm (Foracchia et al., 2005) is also effective for enhancing image regions. But, filters may smooth mineral image details or heavily change the gray distribution. Fuzzy logic based algorithms (Yang et al., 2008; Farbiz et al., 2000) has been used to enhance image, which is also mainly used to remove noises. Utilizing the image information in frequency or wavelet domain performs well for image enhancement in some cases (Agaian et al., 2001; Mencattini et al., 2008). But, the extracted image features in these domains may affect the performance of these algorithms. Morphological operators (Serra, 1982; Soille, 2003), such as top-hat transform (Bai et al., 2012a) and toggle operator (Bai and Zhou, 2011), are also effective tools for image enhancement. Especially, through utilizing the multi-scale theory, image information at different scales could be used for enhancement. Top-hat transform based contrast operator enhances image regions and thus the contrast of an image. But, the image details may not be well enhanced. This will affect the mineral analysis. Toggle operator could be used to enhance details in mineral image, but some details with low contrast may be not well enhanced (Bai and Zhou, 2011; Maragos, 2005; Serra, 1988). To both enhance image contrast and details, sequentially combining the top-hat based contrast operator and toggle contrast operator is an effective way (Bai et al., 2012b) which will produce the resulting image with good contrast and clear image details. This would benefit the further mineral image analysis. Thus, based on the sequential combination of toggle
194
X. Bai / Micron 44 (2013) 193–201
contrast operator and top-hat based contrast operator, an effective mineral image enhancement algorithm could be constructed. In light of this, a mineral image enhancement algorithm through sequentially combining the toggle contrast operator and top-hat based contrast operator is proposed in this paper. Sequential combination of toggle contrast operator and top-hat based contrast operator is discussed, firstly. Then, based on the sequential combination of toggle and top-hat based contrast operator, feature extraction for mineral image enhancement is shown. Moreover, multi-scale extension of the feature extraction is used to extract the multi-scale image features for mineral image enhancement. Through appropriately importing the extracted multi-scale image features into the original mineral image, the mineral image is well enhanced and the image details are clear. The main contributions of this paper are: (1) utilizing the sequentially combined toggle and top-hat based contrast operator for both enhancing mineral image contrast and details; (2) proposing the procedure of feature extraction based on the sequentially combined toggle and top-hat based contrast operator; and (3) extending the feature extraction in multi-scale domain. Experimental results on different mineral images verified the effective performance of the proposed algorithm for mineral image enhancement.
2. Mathematical morphology Mathematical morphology has been well used in wide area of image processing and pattern recognition (Serra, 1982; Soille, 2003). Dilation and erosion are the two basic morphological operators, which are defined as follows. f ⊕ B(x, y) = max(f (x − u, y − v) + B(u, v)), u,v
f B(x, y) = min(f (x + u, y + v) − B(u, v)). u,v
⊕ and represent the dilation and erosion operators, respectively. f represents the mineral image. B is the structuring element. (x, y) and (u, v) are the pixel coordinates of f and B, respectively. Based on dilation and erosion, the morphological opening and closing operators are defined as follows. f ◦ B = (f B) ⊕ B, f • B = (f ⊕ B) B. and 䊉 represent the opening and closing operators, respectively. Opening and closing are useful morphological filters for smoothing bright and dark image regions. Based on opening and closing, the top-hat transforms are defined as follows. WTHB [f (x, y)] = f (x, y) − (f ◦ B)(x, y), BTHB [f (x, y)] = (f • B)(x, y) − f (x, y). WTH and BTH represent the white and black top-hat transforms, respectively. WTH and BTH could be used to extract bright and dark image regions in an image.
3. Mineral image enhancement
widely used toggle contrast operator is defined as follows (Serra, 1982; Soille, 2003).
TCOB [f (x, y)] =
f ⊕ B(x, y),
if
f ⊕ B(x, y) − f (x, y) < f (x, y) − f B(x, y)
f B(x, y),
if
f ⊕ B(x, y) − f (x, y) > f (x, y) − f B(x, y) .
f (x, y),
else
TCO is the selected result from the result of dilation or erosion following the specified rules. Dilation and erosion usually change the marginal regions of image regions. The marginal regions are image details in mineral image. So, TCO contains the identified image details by dilation and erosion, which may be well used for enhancing mineral image details. 3.1.2. Top-hat based contrast operator Top-hat transform extracts bright and dark image regions. By importing the extracted bright and dark image regions into the original image, one top-hat based contrast operator is defined as follows (Serra, 1982; Soille, 2003). THCOB [f (x, y)] = f (x, y) + WTHB [f (x, y)] − BTHB [f (x, y)]. THCO contains bright and dark image regions, and enlarges the contrast between the bright and dark image regions. This enhancement identifies the important image regions. These identified image regions are important regions in mineral image, which could be extracted and well used for mineral image enhancement. The definition of THCO indicates that, THCO operator is not idempotent. Thus, the gray dynamic of the image may be changed. In another way, it is because of the changing of the gray dynamic of the image, THCO could well enhance image regions and the contrast of image. Therefore, THCO is a useful tool for enhancing image contrast. THCO enhances image regions at the scale corresponding to the size of the used structuring element. Although applying THCO several times may apparently enhance image regions at one special scale, the image features at other scales will be suppressed. This may result in the ineffective performance. So, applying THCO several times is not an appropriate way for image enhancement. Instead, using multi-scale structuring elements in THCO, image regions at multi-scales could be enhanced. Then, the contrast of the image which represents the difference between the bright and dark image regions will be effectively enhanced. 3.1.3. Sequential combination of toggle and top-hat based contrast operator TCO identifies the useful mineral image details, and THCO identifies the important mineral image regions. Combining them could both enhance the image details and regions (Bai et al., 2012b) which are important image features in mineral image. These enhanced image features may be extracted for mineral image enhancement. One sequential combination of toggle and top-hat based contrast operator is defined as follows (Bai et al., 2012b). SCOB [f (x, y)] = TCOB [THCOB [f (x, y)]]. In this definition, TCO and THCO are sequentially operated in SCO, which would enhance the important mineral image features, including both the mineral image details and regions. The enhanced image features are identified by the SCO. This would be useful for mineral image enhancement.
3.1. Morphological contrast operator
3.2. Feature extraction
3.1.1. Toggle contrast operator Toggle contrast operator is defined using the primitives and specified rules. Setting dilation and erosion as the primitives, one
The identified image features in SCO have different gray values because of the processing by SCO. Image details include bright and dark image details. Image regions also include bright and dark
X. Bai / Micron 44 (2013) 193–201
image regions. So, the identified image features in SCO include bright and dark image features. The identified bright image features by SCO have larger gray values than the original mineral image, which could be extracted through comparing the gray values of SCO and the original mineral image as follows. BIFB [f (x, y)] = max{SCOB [f (x, y)] − f (x, y), 0}. Similarly, the identified dark image features by SCO have smaller gray values than the original mineral image, which could be extracted through comparing the gray values of SCO and the original mineral image as follows. DIFB [f (x, y)] = max{f (x, y) − SCOB [f (x, y)], 0}. These extracted bright and dark image features could be used for mineral image enhancement. 3.3. Multi-scale feature extraction
could be well enhanced through importing CBIF and CDIF into the original mineral image as follows. FEI(x, y) = f (x, y) + CBIF(x, y) − CDIF(x, y). The bright image features of the original mineral image are enhanced through adding the final extracted bright image features CBIF on the original image. And, the dark image features of the original mineral image are enhanced through subtracting the final extracted dark image features CDIF from the original image. Then, the bright and dark image features in the original mineral image are well enhanced through contrast enlargement. Moreover, the extracted image features by SCO include image details and regions, which means both the image details and regions in the final result mineral image FEI are well enhanced. This would be very useful for the further mineral analysis. 4. Experimental results 4.1. Parameter specification
Structuring element is one important parameter in morphological operators. SCO identifies image features at the scale corresponding to the size of the used structuring element B. Image features exist at different scales of image. To extracted image features at multi-scales of mineral image, multi-scale structuring elements with increasing sizes should be used (Jackway, 1995; Jackway and Deriche, 1996; Jalba et al., 2004). Let B, . . ., nB be the used multi-scale structuring elements which have increasing sizes. sB = B ⊕ B· · · ⊕ B is the structuring
195
dilation s times element corresponding to the scale s, 1 ≤ s ≤ n. By using the multiscale structuring element sB, the bright image features at scale s extracted by SCO could be expressed as follows. BIFsB [f (x, y)] = max{SCOsB [f (x, y)] − f (x, y), 0}. Similarly, the dark image features at scale s extracted by SCO could be expressed as follows. DIFsB [f (x, y)] = max{f (x, y) − SCOsB [f (x, y)], 0}. Through changing the scale number s, the multi-scale bright and dark image features could be well extracted and used for mineral image enhancement. 3.4. Image enhancement The gray values of the pixels of the extracted bright image features at each scale BIFsB are larger than the gray values of the same pixels at other scales. So, the combination of the extracted bright image features at all the scales could be the pixel-wise maximum of BIFsB at all the scales as follows. CBIF = max{BIFsB }. s
Also, the gray values of the pixels of the extracted dark image features at each scale DIFsB are larger than the gray values of the same pixels at other scales. So, the combination of the extracted dark image features at all the scales could be the pixel-wise maximum of DIFsB at all the scales as follows. CDIF = max{DIFsB }. s
CBIF and CDIF are the final extracted bright and dark mineral image features for image enhancement which combines the image features at all the scales of image. So, the original mineral image
The main parameters used in this paper are the structuring element and scale number n. Flat structuring element is one type of simple but widely used structuring element (Jackway and Deriche, 1996; Serra, 1982; Soille, 2003). In this paper, the structuring element is flat structuring element which is mainly decided by the shape and size of the structuring element. Because of the multi-scale extension, the size of the structuring element used in each scale is decided by the scale number. So, the main parameter of structuring element which should be determined is the shape of the structuring element. The circle shape is one widely and effectively used shape in morphological operations. Thus, the shape of the structuring element in this paper is circle. The scale number decides the extracted image features. Larger scale number would extract more image features for mineral image enhancement. However, usually, there is no need to use a very large scale number, and large scale number will dramatically increase the computation time. Our experimental results on different types of mineral images show that, setting n = 3∼5 is effective for most of mineral images. In this paper, we use n = 3. 4.2. Visual result Different types of mineral images are used in this experiment. The used mineral image set is the same as the one used in Bai and Zhou (2011). And, all the experimental results verified the effective performance of the proposed algorithm. Also, to do the comparison, some effective algorithms, including multi-scale toggle operator based algorithm (MST) (Bai and Zhou, 2011), Wallis filter (WF) (Foracchia et al., 2005), histogram equalization algorithm (HE) (Wan and Shi, 2007) and contrast limited adaptive histogram equalization algorithm (CLAHE) (Huang et al., 2006) are used in this paper as the comparison algorithms. HE and CLAHE are widely used image enhancement algorithms, which could be used for mineral image enhancement. WF is the effective filter based image enhancement algorithm, which could be also used for mineral image enhancement. MST is a multi-scale morphology based algorithm for mineral image enhancement. The proposed algorithm is also a multi-scale morphology based algorithm and intends to be an effective algorithm for mineral image enhancement. So, MST, WF, HE and CLAHE are adopted in this paper as the comparison algorithms. Some comparison results are shown below. Fig. 1 shows an example of mineral image with rich details and bright image regions, (a) the original image. The contained rich image details are not clear. (b)–(f) The results of the proposed
196
X. Bai / Micron 44 (2013) 193–201
Fig. 1. Comparison result 1. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
algorithm, MST, WF, HE and CLAHE, respectively. HE and CLAHE enhance the original image. But, the bright image regions are overenhanced, and the image details in these bright regions become even more unclear. WF completely changes the gray distribution of the original image, which will affect the analysis of the minerals. MST and the proposed algorithm well enhance the original image. But, comparing with the result of MST, the result of the proposed algorithm is clearer than the result of MST and the contrast is better. Also, the visible image details in the result of the proposed algorithm are more. So, the proposed algorithm gives the best performance for mineral image enhancement. Fig. 2 shows an example of mineral image containing many unclear image details, (a) the original image. Also, the rich image details are not clear. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. HE and CLAHE do not well enhance the image details and many regions are over enhanced. WF completely changes the gray distribution of the original mineral image. Then, the properties of mineral in image regions with different gray distributions could not be observed. MST well enhances the original mineral image and many image details are clearer than the original image. But, some image details are still not clear. The proposed algorithm gives the best result among these algorithms. Many unclear image details are very clear in the result of the proposed algorithm. And, because the contrast of the result image is good, some invisible image details in the original mineral image could be also clearly observed in the result of the proposed algorithm. The gray distribution of the result is also well maintained. This result would be valuable for mineral analysis. Fig. 3 shows an example of mineral image enhancement with rich image details, (a) the original image. The contained rich image details are also not clear. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. HE enhances the original mineral image, but some image regions are heavily over enhanced. CLAHE enhances the original mineral image, but the image details are still not very clear. Although WF clearly shows the image details, the gray distribution is heavily changed. MST and the proposed algorithm well enhance the original mineral image. Also,
the gray distribution is well maintained by MST and the proposed algorithm. But, the result of the proposed algorithm contains more image details and the image details are clearer. So, the performance of the proposed algorithm is better than other algorithms. Fig. 4 shows an example of mineral image enhancement with many image details and bright image regions, (a) the original mineral image. The contained image details are not clear and there are some bright regions. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. The original mineral image is not clear and some dim image details are also not clear. Because the original mineral image contains some bright image regions, HE and CLAHE over enhance almost all the bright mineral regions, and the image details in these bright regions are still not clear. WF heavily changes the gray distribution of the original mineral image. This may affect the analysis of the processed image. MST and the proposed algorithm well enhance the original mineral image and maintain the gray distribution. Moreover, the proposed algorithm produces an image with rich and clear image details, which gives the best result among these algorithms. Figs. 5 and 6 list some other enhancement results on some more mineral images. In these images, (a) lists the original mineral images and (b) lists the corresponding processed images by the proposed algorithm. The original images contain many unclear image details. Especially, some image details are invisible. After enhancement by the proposed algorithm, the image details are very clear and some invisible image details are visible. Also, the gray distribution is well maintained. This would benefit the further mineral analysis. Different types of mineral images are used in this paper and the results show that the proposed algorithm is effective for mineral image enhancement. Especially, the image details are very clear. 4.3. Quantitative comparison To do a quantitative comparison, the measure spatial frequency (SF) (Aslantas and Kurban, 2009) is used in this paper. The purpose of mineral image enhancement is enhancing the original mineral
X. Bai / Micron 44 (2013) 193–201
197
Fig. 2. Comparison result 2. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
image while making the image details clear and meanwhile maintaining the gray distributions. Thus, the spatial information of the processed image would be good. SF is defined based on the spatial information of an image. Therefore, we use SF as the quantitative measure in this paper. SF of one image I is defined as follows.
SF =
RF 2 + CF 2 ,
where
RF =
CF =
1 [I(x, y) − I(x − 1, y)]2 , M×N M
N
x=1 y=1
1 [I(x, y) − I(x, y − 1)]2 . M×N M
N
x=1 y=1
Fig. 3. Comparison result 3. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
198
X. Bai / Micron 44 (2013) 193–201
Fig. 4. Comparison result 4. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
The definition of SF indicates that, the value of SF is constructed from the difference of the neighboring pixels which could be recognized as the deviation of the gray value of the neighboring pixels. This means, the value of SF is not decided by the dynamic of the
image, which is actually decided by the deviation of the gray value of the neighboring pixels. And, the dynamic of the image may not affect the deviation of the gray value of the neighboring pixels. Moreover, SF has been widely used as one effective measure for
Fig. 5. Some more enhanced images. (a) Original mineral images; and (b) enhanced images.
X. Bai / Micron 44 (2013) 193–201
199
Fig. 6. Some other enhanced images. (a) Original mineral images; and (b) enhanced images.
quantifying the quality of images. Also, SF has been well used as the measure for quantifying the performances of image processing applications, such as image enhancement, image fusion and so on. The proposed mineral image enhancement algorithm could effectively enhance mineral image and image details. Then, the difference information between the neighboring pixels could be well enhanced, which results in mineral image containing more spatial information. SF is defined based on the spatial information of an image represented by the difference information between the neighboring pixels. Therefore, using SF as the quantitative measure in this paper is appropriate. A bigger value of SF indicates that, the performance of the corresponding mineral image enhancement algorithm is better. HE, CLAHE, WF, MST and the proposed algorithm are applied on different mineral images and the processed images are used to calculate the SF value. The SF values of the original and processed mineral images of Figs. 1–4 are shown in Fig. 7. In Fig. 7, (a) is the quantitative comparison on images shown in Fig. 1. (b) Quantitative comparison on images shown in Fig. 2. (c) Quantitative comparison on images shown in Fig. 3. (d) Quantitative comparison on images shown in Fig. 4. These quantitative comparisons show that, the SF values of the proposed algorithm are larger than other algorithms in most of the cases. This means, the performances of the proposed algorithm for enhancing mineral images and making image details clear are better than other algorithms. Therefore,
the proposed algorithm is effective for mineral image enhancement. To give an overall quantitative comparison, the mean value of the SF values of all the processed images by each algorithm is shown in Fig. 8. Fig. 8 shows that, the proposed algorithm gets the biggest SF value comparing with other algorithms. This indicates that, the proposed algorithm could well enhance mineral images and make image details clear. Moreover, the used mineral images are different microscopy images. And, the proposed algorithm performs well on these images. It should be noticed that, because no special property of microscopes is needed, the proposed algorithm could be used for enhancing different types of microscopy images obtained from different microscope modalities. Moreover, the proposed algorithm may be also extended to enhance other types of images with low contrast and rich image details. Moreover, because of the multi-scale theory, the operators are calculated several times. This may increase the calculation time of the proposed algorithm. However, morphological operators have accelerating strategies (Droogenbroeck and Talbot, 1996; Gonzalez et al., 2002; Park and Chin, 1995) to speed up the calculation of the operators, such as structuring element decomposition, parallel calculation and so on. By using these strategies, the proposed algorithm may be performed fast. All of these indicate that, the proposed algorithm could be effectively used as a tool for mineral analysis, which may be performed effectively and fast. Also, it is applicable for different microscope modalities.
200
X. Bai / Micron 44 (2013) 193–201
Fig. 7. Quantitative comparisons using the measure SF on images shown in Figs. 1–4. (a) Quantitative comparison on images shown in Fig. 1; (b) quantitative comparison on images shown in Fig. 2; (c) quantitative comparison on images shown in Fig. 3; and (d) quantitative comparison on images shown in Fig. 4.
Fig. 8. Overall quantitative comparison using the measure SF.
5. Conclusions Enhancing mineral image and making image details clear are crucial techniques for analyzing mineral properties. The very important points of mineral image enhancement are making image details clear and meanwhile maintaining the gray distributions, so that the properties of minerals could be well maintained and clear for the further analysis. These make the mineral analysis easy and effective. Morphological operators are useful tool for image analysis. And, the sequentially combining of morphological toggle contrast operator and top-hat based contrast operator could well extract mineral image features and makes image details clear. Also, the gray distribution could be well maintained.
Therefore, based on the sequentially combining of toggle contrast operator and top-hat based contrast operator, the algorithm proposed in this paper is a useful tool for mineral image enhancement. Experiments on different types of mineral images show that, the performance of the proposed algorithm for mineral image enhancement is effective and better than some other algorithms. Because of the effective mineral feature extraction, both the contrast and image details are well enhanced. Also, because of the multi-scale extension of feature extraction, the useful mineral image features are well explored for the further mineral analysis. Therefore, the proposed algorithm could be well used for mineral analysis. Moreover, the proposed algorithm does not need special property of microscopes and the morphological operators have accelerating strategies. Therefore, the proposed algorithm could be used for different types of microscopy images and may be used as an efficient tool for the further mineral analysis.
Acknowledgments The author thanks the anonymous reviewers for their valuable comments and suggestions. This work has been partly supported by the National Natural Science Foundation of China (Grant No. 60902056), open funding project of State Key Laboratory of Virtual Reality Technology and Systems, Beihang University (Grant No. BUAA-VR-12KF-04), Fundamental Research Funds for the Central Universities (Grant No. YWF-11-03-Q-065) and Beijing Key Laboratory of Digital Media.
X. Bai / Micron 44 (2013) 193–201
References Agaian, S.S., Panetta, K., Grigoryan, A.M., 2001. Transform-based image enhancement algorithms with performance measure. IEEE Transactions on Image Processing 10, 267–382. Aslantas, V., Kurban, R., 2009. A comparison of criterion functions for fusion of multifocus noisy images. Optics Communications 282, 3231–3242. Bai, X., Zhou, F., 2011. Multi scale toggle contrast operator based mineral image enhancement. Journal of Microscopy 243, 141–153. Bai, X., Zhou, F., Xue, B., 2012a. Image enhancement using multi scale image features extracted by top-hat transform. Optics & Laser Technology 44, 328–336. Bai, X., Zhou, F., Xue, B., 2012b. Toggle and top-hat based morphological contrast operators. Computers & Electrical Engineering, http://dx.doi.org/10.1016/j.compeleceng.2011.08.010. Carter, D.H., Scully, A.J., Hatton, P.V., Davies, R.M., Aaron, J.E., 2000. Cryopreservation and image enhancement of juvenile and adult dentine mineral. The Histochemical Journal 32, 253–261. Droogenbroeck, M., Talbot, H., 1996. Fast computation of morphological operations with arbitrary structuring elements. Pattern Recognition Letters 17, 1451–1460. Farbiz, F., Menhaj, M.B., Motamedi, S.A., Hagan, M.T., 2000. A new fuzzy logic filter for image enhancement. IEEE Transactions on Systems, Man, and Cybernetics – Part B: Cybernetics 30, 110–119. Foracchia, M., Grisan, E., Ruggeri, A., 2005. Luminosity and contrast normalization in retinal images. Medical Image Analysis 9, 179–190. Gilboa, G., Sochen, N., Zeevi, Y.Y., 2004. Image enhancement and denoising by complex diffusion processes. IEEE Transactions on Pattern Analysis and Machine Intelligence 26, 1020–1036. Gonzalez, F., Tubıo, O., Tobajas, F., Armas, V., Esper-Chaın, R., Sarmiento, R., 2002. Morphological processor for real-time image applications. Microelectronics Journal 33, 1115–1122. Heintzmann, R., Sarafis, V., Munroe, P., Nailon, J., Hanley, Q.S., Jovin, T.M., 2003. Resolution enhancement by subtraction of confocal signals taken at different pinhole sizes. Micron 34, 293–300. Huang, K.Q., Wang, Q., Wu, Z.Y., 2006. Natural color image enhancement and evaluation algorithm based on human visual system. Computer Vision and Image Understanding 103, 52–63. Jackway, P., 1995. On dimensionality in multiscale morphological scale-space with elliptical poweroid structuring functions. Journal of Visual Communication and Image Representation 6, 189–195. Jackway, P., Deriche, M., 1996. Scale-space properties of the multiscale morphological dilation-erosion. IEEE Transactions on Pattern Analysis and Machine Intelligence 18, 38–51.
201
Jalba, A., Wilkinson, M., Roerdink, J., 2004. Morphological hat-transform scale spaces and their use in pattern classification. Pattern Recognition 37, 901–915. Jones, A.S., Reztsov, A., Loo, C.E., 2007. Application of invariant grey scale features for analysis of porous minerals. Micron 38, 40–48. Kaiser, U., Chuvilin, A., 2003. Enhanced compositional contrast in imaging of nanoprecipitates buried in a defective crystal using a conventional TEM. Microscopy and Microanalysis 9, 36–41. Lumpkin, G.R., Smith, K.L., Giere, R., 1997. Application of analytical electron microscopy to the study of radiation damage in the complex oxide mineral zirconolite. Micron 28, 57–68. Maragos, P., 2005. Morphological filtering for image enhancement and feature detection. In: Bovik, A.C. (Ed.), The Image and Video Processing Handbook. , 2nd ed. Elsevier Academic Press, pp. 135–156. Mencattini, A., Salmeri, M., Lojacono, R., Frigerio, M., Caselli, F., 2008. Mammographic images enhancement and denoising for breast cancer detection using dyadic wavelet processing. IEEE Transactions on Instrumentation and Measurement 57, 1422–1430. Park, H., Chin, R., 1995. Decomposition of arbitrarily shaped morphological structuring elements. IEEE Transactions on Pattern Analysis and Machine Intelligence 17, 2–15. Reshak, A.H., 2009. Enhancing the resolution of the forward second harmonic imaging using the two-photon laser scanning microscope. Micron 40, 750–755. Serra, J., 1982. Image Analysis and Mathematical Morphology. Academic Press, New York. Serra, J., 1988. Toggle Mappings. Technical Report N-18/88/MM, Centre de Morphologie Mathematique, ENSMP, Fontainebleau, France. Schilders, S.P., Gan, X.S., Gu, M., 1998. Image enhancement in a reflection optical microscope by suppression of diffusing photons using polarizing annular objectives. Microscopy and Microanalysis 4, 415–419. Schultze-Lam, S., Harauz, G., Beveridge, T.J., 1992. Participation of a cyanobacterial S layer in fine-grain mineral formation. Journal of Bacteriology 174, 7971–7981. Shah, S., 2007. Image enhancement for increased dot-counting efficiency in FISH. Journal of Microscopy 228, 211–226. Soille, P., 2003. Morphological Image Analysis – Principle and Applications. Springer, Germany. Tang, B., Sapiro, G., Caselles, V., 2001. Color image enhancement via chromaticity diffusion. IEEE Transactions on Image Processing 10, 701–707. Wan, Y., Shi, D., 2007. Joint exact histogram specification and image enhancement through the wavelet transform. IEEE Transactions on Image Processing 16, 2245–2250. Yang, C., Lu, L., Lin, H., Guan, R., Shi, X., Liang, Y., 2008. A fuzzy-statisticsbased principal component analysis (FS-PCA) method for multispectral image enhancement and display. IEEE Transactions on Geoscience and Remote Sensing 46, 3937–3947.
Contents lists available at SciVerse ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
Mineral image enhancement based on sequential combination of toggle and top-hat based contrast operator Xiangzhi Bai ∗ Image Processing Centre, Beijing University of Aeronautics and Astronautics, 100191 Beijing, China
a r t i c l e
i n f o
Article history: Received 13 March 2012 Received in revised form 7 June 2012 Accepted 17 June 2012 Keywords: Toggle contrast operator Top-hat based contrast operator Mineral image enhancement Mathematical morphology
a b s t r a c t Enhancing mineral image especially making mineral image details clear is very useful for mineral analysis. To effectively enhance mineral image, an algorithm based on the toggle contrast operator and top-hat based contrast operator is proposed in this paper. Sequentially combining the toggle contrast operator and top-hat based contrast operator could be used to identify image features especially the image details. So, appropriately exacting the identified image features by the sequentially combined toggle and tophat based contrast operator is important for mineral image enhancement, which is analyzed firstly in this paper. After that, the multi-scale extension of feature extraction is given and used to construct the final features for mineral image enhancement. By importing the final extracted image features into the original mineral image through contrast enlargement, the original mineral image is well enhanced and the mineral image details are very clear. Experimental results on different types of mineral images verified the effective performance of the proposed algorithm. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Microscopy is the important instrument for mineral analysis (Kaiser and Chuvilin, 2003; Shah, 2007; Lumpkin et al., 1997). Obtained mineral images from different types of microscopes, such as scanning electron microscopy, scanning tunneling microscopy and transmission electron microscopy, could provide valuable information for mineral property analysis, new mineral identification, interactions analysis of different minerals and so on (Schultze-Lam et al., 1992; Carter et al., 2000; Jones et al., 2007). Designing microscopy is an effective way to obtain clear images (Schilders et al., 1998; Reshak, 2009). However, because the operators of microscopy equipment may not well produce the image of mineral samples or the imaging procedure is too difficult to produce clear mineral images, many obtained mineral images do not have good contrast and the contained details are usually not clear. These may affect the applications of these mineral images for the further mineral analysis. Therefore, it is useful to well enhance the mineral images and makes the image details clear. Image enhancement technique, which is an active research area in image processing, is the effective way for mineral image enhancement (Bai and Zhou, 2011; Heintzmann et al., 2003). Histogram based algorithms (Huang et al., 2006; Wan and Shi, 2007) are widely used and effective ways for image enhancement. But,
∗ Tel.: +86 1082338578. E-mail address: [email protected] 0968-4328/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.micron.2012.06.009
the bright image regions in mineral image may be over enhanced. Diffusion based algorithms (Tang et al., 2001; Gilboa et al., 2004) are mainly focusing on noise reduction, which is useful for noise smoothing. Filter based algorithm (Foracchia et al., 2005) is also effective for enhancing image regions. But, filters may smooth mineral image details or heavily change the gray distribution. Fuzzy logic based algorithms (Yang et al., 2008; Farbiz et al., 2000) has been used to enhance image, which is also mainly used to remove noises. Utilizing the image information in frequency or wavelet domain performs well for image enhancement in some cases (Agaian et al., 2001; Mencattini et al., 2008). But, the extracted image features in these domains may affect the performance of these algorithms. Morphological operators (Serra, 1982; Soille, 2003), such as top-hat transform (Bai et al., 2012a) and toggle operator (Bai and Zhou, 2011), are also effective tools for image enhancement. Especially, through utilizing the multi-scale theory, image information at different scales could be used for enhancement. Top-hat transform based contrast operator enhances image regions and thus the contrast of an image. But, the image details may not be well enhanced. This will affect the mineral analysis. Toggle operator could be used to enhance details in mineral image, but some details with low contrast may be not well enhanced (Bai and Zhou, 2011; Maragos, 2005; Serra, 1988). To both enhance image contrast and details, sequentially combining the top-hat based contrast operator and toggle contrast operator is an effective way (Bai et al., 2012b) which will produce the resulting image with good contrast and clear image details. This would benefit the further mineral image analysis. Thus, based on the sequential combination of toggle
194
X. Bai / Micron 44 (2013) 193–201
contrast operator and top-hat based contrast operator, an effective mineral image enhancement algorithm could be constructed. In light of this, a mineral image enhancement algorithm through sequentially combining the toggle contrast operator and top-hat based contrast operator is proposed in this paper. Sequential combination of toggle contrast operator and top-hat based contrast operator is discussed, firstly. Then, based on the sequential combination of toggle and top-hat based contrast operator, feature extraction for mineral image enhancement is shown. Moreover, multi-scale extension of the feature extraction is used to extract the multi-scale image features for mineral image enhancement. Through appropriately importing the extracted multi-scale image features into the original mineral image, the mineral image is well enhanced and the image details are clear. The main contributions of this paper are: (1) utilizing the sequentially combined toggle and top-hat based contrast operator for both enhancing mineral image contrast and details; (2) proposing the procedure of feature extraction based on the sequentially combined toggle and top-hat based contrast operator; and (3) extending the feature extraction in multi-scale domain. Experimental results on different mineral images verified the effective performance of the proposed algorithm for mineral image enhancement.
2. Mathematical morphology Mathematical morphology has been well used in wide area of image processing and pattern recognition (Serra, 1982; Soille, 2003). Dilation and erosion are the two basic morphological operators, which are defined as follows. f ⊕ B(x, y) = max(f (x − u, y − v) + B(u, v)), u,v
f B(x, y) = min(f (x + u, y + v) − B(u, v)). u,v
⊕ and represent the dilation and erosion operators, respectively. f represents the mineral image. B is the structuring element. (x, y) and (u, v) are the pixel coordinates of f and B, respectively. Based on dilation and erosion, the morphological opening and closing operators are defined as follows. f ◦ B = (f B) ⊕ B, f • B = (f ⊕ B) B. and 䊉 represent the opening and closing operators, respectively. Opening and closing are useful morphological filters for smoothing bright and dark image regions. Based on opening and closing, the top-hat transforms are defined as follows. WTHB [f (x, y)] = f (x, y) − (f ◦ B)(x, y), BTHB [f (x, y)] = (f • B)(x, y) − f (x, y). WTH and BTH represent the white and black top-hat transforms, respectively. WTH and BTH could be used to extract bright and dark image regions in an image.
3. Mineral image enhancement
widely used toggle contrast operator is defined as follows (Serra, 1982; Soille, 2003).
TCOB [f (x, y)] =
f ⊕ B(x, y),
if
f ⊕ B(x, y) − f (x, y) < f (x, y) − f B(x, y)
f B(x, y),
if
f ⊕ B(x, y) − f (x, y) > f (x, y) − f B(x, y) .
f (x, y),
else
TCO is the selected result from the result of dilation or erosion following the specified rules. Dilation and erosion usually change the marginal regions of image regions. The marginal regions are image details in mineral image. So, TCO contains the identified image details by dilation and erosion, which may be well used for enhancing mineral image details. 3.1.2. Top-hat based contrast operator Top-hat transform extracts bright and dark image regions. By importing the extracted bright and dark image regions into the original image, one top-hat based contrast operator is defined as follows (Serra, 1982; Soille, 2003). THCOB [f (x, y)] = f (x, y) + WTHB [f (x, y)] − BTHB [f (x, y)]. THCO contains bright and dark image regions, and enlarges the contrast between the bright and dark image regions. This enhancement identifies the important image regions. These identified image regions are important regions in mineral image, which could be extracted and well used for mineral image enhancement. The definition of THCO indicates that, THCO operator is not idempotent. Thus, the gray dynamic of the image may be changed. In another way, it is because of the changing of the gray dynamic of the image, THCO could well enhance image regions and the contrast of image. Therefore, THCO is a useful tool for enhancing image contrast. THCO enhances image regions at the scale corresponding to the size of the used structuring element. Although applying THCO several times may apparently enhance image regions at one special scale, the image features at other scales will be suppressed. This may result in the ineffective performance. So, applying THCO several times is not an appropriate way for image enhancement. Instead, using multi-scale structuring elements in THCO, image regions at multi-scales could be enhanced. Then, the contrast of the image which represents the difference between the bright and dark image regions will be effectively enhanced. 3.1.3. Sequential combination of toggle and top-hat based contrast operator TCO identifies the useful mineral image details, and THCO identifies the important mineral image regions. Combining them could both enhance the image details and regions (Bai et al., 2012b) which are important image features in mineral image. These enhanced image features may be extracted for mineral image enhancement. One sequential combination of toggle and top-hat based contrast operator is defined as follows (Bai et al., 2012b). SCOB [f (x, y)] = TCOB [THCOB [f (x, y)]]. In this definition, TCO and THCO are sequentially operated in SCO, which would enhance the important mineral image features, including both the mineral image details and regions. The enhanced image features are identified by the SCO. This would be useful for mineral image enhancement.
3.1. Morphological contrast operator
3.2. Feature extraction
3.1.1. Toggle contrast operator Toggle contrast operator is defined using the primitives and specified rules. Setting dilation and erosion as the primitives, one
The identified image features in SCO have different gray values because of the processing by SCO. Image details include bright and dark image details. Image regions also include bright and dark
X. Bai / Micron 44 (2013) 193–201
image regions. So, the identified image features in SCO include bright and dark image features. The identified bright image features by SCO have larger gray values than the original mineral image, which could be extracted through comparing the gray values of SCO and the original mineral image as follows. BIFB [f (x, y)] = max{SCOB [f (x, y)] − f (x, y), 0}. Similarly, the identified dark image features by SCO have smaller gray values than the original mineral image, which could be extracted through comparing the gray values of SCO and the original mineral image as follows. DIFB [f (x, y)] = max{f (x, y) − SCOB [f (x, y)], 0}. These extracted bright and dark image features could be used for mineral image enhancement. 3.3. Multi-scale feature extraction
could be well enhanced through importing CBIF and CDIF into the original mineral image as follows. FEI(x, y) = f (x, y) + CBIF(x, y) − CDIF(x, y). The bright image features of the original mineral image are enhanced through adding the final extracted bright image features CBIF on the original image. And, the dark image features of the original mineral image are enhanced through subtracting the final extracted dark image features CDIF from the original image. Then, the bright and dark image features in the original mineral image are well enhanced through contrast enlargement. Moreover, the extracted image features by SCO include image details and regions, which means both the image details and regions in the final result mineral image FEI are well enhanced. This would be very useful for the further mineral analysis. 4. Experimental results 4.1. Parameter specification
Structuring element is one important parameter in morphological operators. SCO identifies image features at the scale corresponding to the size of the used structuring element B. Image features exist at different scales of image. To extracted image features at multi-scales of mineral image, multi-scale structuring elements with increasing sizes should be used (Jackway, 1995; Jackway and Deriche, 1996; Jalba et al., 2004). Let B, . . ., nB be the used multi-scale structuring elements which have increasing sizes. sB = B ⊕ B· · · ⊕ B is the structuring
195
dilation s times element corresponding to the scale s, 1 ≤ s ≤ n. By using the multiscale structuring element sB, the bright image features at scale s extracted by SCO could be expressed as follows. BIFsB [f (x, y)] = max{SCOsB [f (x, y)] − f (x, y), 0}. Similarly, the dark image features at scale s extracted by SCO could be expressed as follows. DIFsB [f (x, y)] = max{f (x, y) − SCOsB [f (x, y)], 0}. Through changing the scale number s, the multi-scale bright and dark image features could be well extracted and used for mineral image enhancement. 3.4. Image enhancement The gray values of the pixels of the extracted bright image features at each scale BIFsB are larger than the gray values of the same pixels at other scales. So, the combination of the extracted bright image features at all the scales could be the pixel-wise maximum of BIFsB at all the scales as follows. CBIF = max{BIFsB }. s
Also, the gray values of the pixels of the extracted dark image features at each scale DIFsB are larger than the gray values of the same pixels at other scales. So, the combination of the extracted dark image features at all the scales could be the pixel-wise maximum of DIFsB at all the scales as follows. CDIF = max{DIFsB }. s
CBIF and CDIF are the final extracted bright and dark mineral image features for image enhancement which combines the image features at all the scales of image. So, the original mineral image
The main parameters used in this paper are the structuring element and scale number n. Flat structuring element is one type of simple but widely used structuring element (Jackway and Deriche, 1996; Serra, 1982; Soille, 2003). In this paper, the structuring element is flat structuring element which is mainly decided by the shape and size of the structuring element. Because of the multi-scale extension, the size of the structuring element used in each scale is decided by the scale number. So, the main parameter of structuring element which should be determined is the shape of the structuring element. The circle shape is one widely and effectively used shape in morphological operations. Thus, the shape of the structuring element in this paper is circle. The scale number decides the extracted image features. Larger scale number would extract more image features for mineral image enhancement. However, usually, there is no need to use a very large scale number, and large scale number will dramatically increase the computation time. Our experimental results on different types of mineral images show that, setting n = 3∼5 is effective for most of mineral images. In this paper, we use n = 3. 4.2. Visual result Different types of mineral images are used in this experiment. The used mineral image set is the same as the one used in Bai and Zhou (2011). And, all the experimental results verified the effective performance of the proposed algorithm. Also, to do the comparison, some effective algorithms, including multi-scale toggle operator based algorithm (MST) (Bai and Zhou, 2011), Wallis filter (WF) (Foracchia et al., 2005), histogram equalization algorithm (HE) (Wan and Shi, 2007) and contrast limited adaptive histogram equalization algorithm (CLAHE) (Huang et al., 2006) are used in this paper as the comparison algorithms. HE and CLAHE are widely used image enhancement algorithms, which could be used for mineral image enhancement. WF is the effective filter based image enhancement algorithm, which could be also used for mineral image enhancement. MST is a multi-scale morphology based algorithm for mineral image enhancement. The proposed algorithm is also a multi-scale morphology based algorithm and intends to be an effective algorithm for mineral image enhancement. So, MST, WF, HE and CLAHE are adopted in this paper as the comparison algorithms. Some comparison results are shown below. Fig. 1 shows an example of mineral image with rich details and bright image regions, (a) the original image. The contained rich image details are not clear. (b)–(f) The results of the proposed
196
X. Bai / Micron 44 (2013) 193–201
Fig. 1. Comparison result 1. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
algorithm, MST, WF, HE and CLAHE, respectively. HE and CLAHE enhance the original image. But, the bright image regions are overenhanced, and the image details in these bright regions become even more unclear. WF completely changes the gray distribution of the original image, which will affect the analysis of the minerals. MST and the proposed algorithm well enhance the original image. But, comparing with the result of MST, the result of the proposed algorithm is clearer than the result of MST and the contrast is better. Also, the visible image details in the result of the proposed algorithm are more. So, the proposed algorithm gives the best performance for mineral image enhancement. Fig. 2 shows an example of mineral image containing many unclear image details, (a) the original image. Also, the rich image details are not clear. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. HE and CLAHE do not well enhance the image details and many regions are over enhanced. WF completely changes the gray distribution of the original mineral image. Then, the properties of mineral in image regions with different gray distributions could not be observed. MST well enhances the original mineral image and many image details are clearer than the original image. But, some image details are still not clear. The proposed algorithm gives the best result among these algorithms. Many unclear image details are very clear in the result of the proposed algorithm. And, because the contrast of the result image is good, some invisible image details in the original mineral image could be also clearly observed in the result of the proposed algorithm. The gray distribution of the result is also well maintained. This result would be valuable for mineral analysis. Fig. 3 shows an example of mineral image enhancement with rich image details, (a) the original image. The contained rich image details are also not clear. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. HE enhances the original mineral image, but some image regions are heavily over enhanced. CLAHE enhances the original mineral image, but the image details are still not very clear. Although WF clearly shows the image details, the gray distribution is heavily changed. MST and the proposed algorithm well enhance the original mineral image. Also,
the gray distribution is well maintained by MST and the proposed algorithm. But, the result of the proposed algorithm contains more image details and the image details are clearer. So, the performance of the proposed algorithm is better than other algorithms. Fig. 4 shows an example of mineral image enhancement with many image details and bright image regions, (a) the original mineral image. The contained image details are not clear and there are some bright regions. (b)–(f) The results of the proposed algorithm, MST, WF, HE and CLAHE, respectively. The original mineral image is not clear and some dim image details are also not clear. Because the original mineral image contains some bright image regions, HE and CLAHE over enhance almost all the bright mineral regions, and the image details in these bright regions are still not clear. WF heavily changes the gray distribution of the original mineral image. This may affect the analysis of the processed image. MST and the proposed algorithm well enhance the original mineral image and maintain the gray distribution. Moreover, the proposed algorithm produces an image with rich and clear image details, which gives the best result among these algorithms. Figs. 5 and 6 list some other enhancement results on some more mineral images. In these images, (a) lists the original mineral images and (b) lists the corresponding processed images by the proposed algorithm. The original images contain many unclear image details. Especially, some image details are invisible. After enhancement by the proposed algorithm, the image details are very clear and some invisible image details are visible. Also, the gray distribution is well maintained. This would benefit the further mineral analysis. Different types of mineral images are used in this paper and the results show that the proposed algorithm is effective for mineral image enhancement. Especially, the image details are very clear. 4.3. Quantitative comparison To do a quantitative comparison, the measure spatial frequency (SF) (Aslantas and Kurban, 2009) is used in this paper. The purpose of mineral image enhancement is enhancing the original mineral
X. Bai / Micron 44 (2013) 193–201
197
Fig. 2. Comparison result 2. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
image while making the image details clear and meanwhile maintaining the gray distributions. Thus, the spatial information of the processed image would be good. SF is defined based on the spatial information of an image. Therefore, we use SF as the quantitative measure in this paper. SF of one image I is defined as follows.
SF =
RF 2 + CF 2 ,
where
RF =
CF =
1 [I(x, y) − I(x − 1, y)]2 , M×N M
N
x=1 y=1
1 [I(x, y) − I(x, y − 1)]2 . M×N M
N
x=1 y=1
Fig. 3. Comparison result 3. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
198
X. Bai / Micron 44 (2013) 193–201
Fig. 4. Comparison result 4. (a) Original image; (b) enhanced result by the proposed algorithm; (c) enhanced result by MST; (d) enhanced result by WF; (e) enhanced result by HE; and (f) enhanced result by CLAHE.
The definition of SF indicates that, the value of SF is constructed from the difference of the neighboring pixels which could be recognized as the deviation of the gray value of the neighboring pixels. This means, the value of SF is not decided by the dynamic of the
image, which is actually decided by the deviation of the gray value of the neighboring pixels. And, the dynamic of the image may not affect the deviation of the gray value of the neighboring pixels. Moreover, SF has been widely used as one effective measure for
Fig. 5. Some more enhanced images. (a) Original mineral images; and (b) enhanced images.
X. Bai / Micron 44 (2013) 193–201
199
Fig. 6. Some other enhanced images. (a) Original mineral images; and (b) enhanced images.
quantifying the quality of images. Also, SF has been well used as the measure for quantifying the performances of image processing applications, such as image enhancement, image fusion and so on. The proposed mineral image enhancement algorithm could effectively enhance mineral image and image details. Then, the difference information between the neighboring pixels could be well enhanced, which results in mineral image containing more spatial information. SF is defined based on the spatial information of an image represented by the difference information between the neighboring pixels. Therefore, using SF as the quantitative measure in this paper is appropriate. A bigger value of SF indicates that, the performance of the corresponding mineral image enhancement algorithm is better. HE, CLAHE, WF, MST and the proposed algorithm are applied on different mineral images and the processed images are used to calculate the SF value. The SF values of the original and processed mineral images of Figs. 1–4 are shown in Fig. 7. In Fig. 7, (a) is the quantitative comparison on images shown in Fig. 1. (b) Quantitative comparison on images shown in Fig. 2. (c) Quantitative comparison on images shown in Fig. 3. (d) Quantitative comparison on images shown in Fig. 4. These quantitative comparisons show that, the SF values of the proposed algorithm are larger than other algorithms in most of the cases. This means, the performances of the proposed algorithm for enhancing mineral images and making image details clear are better than other algorithms. Therefore,
the proposed algorithm is effective for mineral image enhancement. To give an overall quantitative comparison, the mean value of the SF values of all the processed images by each algorithm is shown in Fig. 8. Fig. 8 shows that, the proposed algorithm gets the biggest SF value comparing with other algorithms. This indicates that, the proposed algorithm could well enhance mineral images and make image details clear. Moreover, the used mineral images are different microscopy images. And, the proposed algorithm performs well on these images. It should be noticed that, because no special property of microscopes is needed, the proposed algorithm could be used for enhancing different types of microscopy images obtained from different microscope modalities. Moreover, the proposed algorithm may be also extended to enhance other types of images with low contrast and rich image details. Moreover, because of the multi-scale theory, the operators are calculated several times. This may increase the calculation time of the proposed algorithm. However, morphological operators have accelerating strategies (Droogenbroeck and Talbot, 1996; Gonzalez et al., 2002; Park and Chin, 1995) to speed up the calculation of the operators, such as structuring element decomposition, parallel calculation and so on. By using these strategies, the proposed algorithm may be performed fast. All of these indicate that, the proposed algorithm could be effectively used as a tool for mineral analysis, which may be performed effectively and fast. Also, it is applicable for different microscope modalities.
200
X. Bai / Micron 44 (2013) 193–201
Fig. 7. Quantitative comparisons using the measure SF on images shown in Figs. 1–4. (a) Quantitative comparison on images shown in Fig. 1; (b) quantitative comparison on images shown in Fig. 2; (c) quantitative comparison on images shown in Fig. 3; and (d) quantitative comparison on images shown in Fig. 4.
Fig. 8. Overall quantitative comparison using the measure SF.
5. Conclusions Enhancing mineral image and making image details clear are crucial techniques for analyzing mineral properties. The very important points of mineral image enhancement are making image details clear and meanwhile maintaining the gray distributions, so that the properties of minerals could be well maintained and clear for the further analysis. These make the mineral analysis easy and effective. Morphological operators are useful tool for image analysis. And, the sequentially combining of morphological toggle contrast operator and top-hat based contrast operator could well extract mineral image features and makes image details clear. Also, the gray distribution could be well maintained.
Therefore, based on the sequentially combining of toggle contrast operator and top-hat based contrast operator, the algorithm proposed in this paper is a useful tool for mineral image enhancement. Experiments on different types of mineral images show that, the performance of the proposed algorithm for mineral image enhancement is effective and better than some other algorithms. Because of the effective mineral feature extraction, both the contrast and image details are well enhanced. Also, because of the multi-scale extension of feature extraction, the useful mineral image features are well explored for the further mineral analysis. Therefore, the proposed algorithm could be well used for mineral analysis. Moreover, the proposed algorithm does not need special property of microscopes and the morphological operators have accelerating strategies. Therefore, the proposed algorithm could be used for different types of microscopy images and may be used as an efficient tool for the further mineral analysis.
Acknowledgments The author thanks the anonymous reviewers for their valuable comments and suggestions. This work has been partly supported by the National Natural Science Foundation of China (Grant No. 60902056), open funding project of State Key Laboratory of Virtual Reality Technology and Systems, Beihang University (Grant No. BUAA-VR-12KF-04), Fundamental Research Funds for the Central Universities (Grant No. YWF-11-03-Q-065) and Beijing Key Laboratory of Digital Media.
X. Bai / Micron 44 (2013) 193–201
References Agaian, S.S., Panetta, K., Grigoryan, A.M., 2001. Transform-based image enhancement algorithms with performance measure. IEEE Transactions on Image Processing 10, 267–382. Aslantas, V., Kurban, R., 2009. A comparison of criterion functions for fusion of multifocus noisy images. Optics Communications 282, 3231–3242. Bai, X., Zhou, F., 2011. Multi scale toggle contrast operator based mineral image enhancement. Journal of Microscopy 243, 141–153. Bai, X., Zhou, F., Xue, B., 2012a. Image enhancement using multi scale image features extracted by top-hat transform. Optics & Laser Technology 44, 328–336. Bai, X., Zhou, F., Xue, B., 2012b. Toggle and top-hat based morphological contrast operators. Computers & Electrical Engineering, http://dx.doi.org/10.1016/j.compeleceng.2011.08.010. Carter, D.H., Scully, A.J., Hatton, P.V., Davies, R.M., Aaron, J.E., 2000. Cryopreservation and image enhancement of juvenile and adult dentine mineral. The Histochemical Journal 32, 253–261. Droogenbroeck, M., Talbot, H., 1996. Fast computation of morphological operations with arbitrary structuring elements. Pattern Recognition Letters 17, 1451–1460. Farbiz, F., Menhaj, M.B., Motamedi, S.A., Hagan, M.T., 2000. A new fuzzy logic filter for image enhancement. IEEE Transactions on Systems, Man, and Cybernetics – Part B: Cybernetics 30, 110–119. Foracchia, M., Grisan, E., Ruggeri, A., 2005. Luminosity and contrast normalization in retinal images. Medical Image Analysis 9, 179–190. Gilboa, G., Sochen, N., Zeevi, Y.Y., 2004. Image enhancement and denoising by complex diffusion processes. IEEE Transactions on Pattern Analysis and Machine Intelligence 26, 1020–1036. Gonzalez, F., Tubıo, O., Tobajas, F., Armas, V., Esper-Chaın, R., Sarmiento, R., 2002. Morphological processor for real-time image applications. Microelectronics Journal 33, 1115–1122. Heintzmann, R., Sarafis, V., Munroe, P., Nailon, J., Hanley, Q.S., Jovin, T.M., 2003. Resolution enhancement by subtraction of confocal signals taken at different pinhole sizes. Micron 34, 293–300. Huang, K.Q., Wang, Q., Wu, Z.Y., 2006. Natural color image enhancement and evaluation algorithm based on human visual system. Computer Vision and Image Understanding 103, 52–63. Jackway, P., 1995. On dimensionality in multiscale morphological scale-space with elliptical poweroid structuring functions. Journal of Visual Communication and Image Representation 6, 189–195. Jackway, P., Deriche, M., 1996. Scale-space properties of the multiscale morphological dilation-erosion. IEEE Transactions on Pattern Analysis and Machine Intelligence 18, 38–51.
201
Jalba, A., Wilkinson, M., Roerdink, J., 2004. Morphological hat-transform scale spaces and their use in pattern classification. Pattern Recognition 37, 901–915. Jones, A.S., Reztsov, A., Loo, C.E., 2007. Application of invariant grey scale features for analysis of porous minerals. Micron 38, 40–48. Kaiser, U., Chuvilin, A., 2003. Enhanced compositional contrast in imaging of nanoprecipitates buried in a defective crystal using a conventional TEM. Microscopy and Microanalysis 9, 36–41. Lumpkin, G.R., Smith, K.L., Giere, R., 1997. Application of analytical electron microscopy to the study of radiation damage in the complex oxide mineral zirconolite. Micron 28, 57–68. Maragos, P., 2005. Morphological filtering for image enhancement and feature detection. In: Bovik, A.C. (Ed.), The Image and Video Processing Handbook. , 2nd ed. Elsevier Academic Press, pp. 135–156. Mencattini, A., Salmeri, M., Lojacono, R., Frigerio, M., Caselli, F., 2008. Mammographic images enhancement and denoising for breast cancer detection using dyadic wavelet processing. IEEE Transactions on Instrumentation and Measurement 57, 1422–1430. Park, H., Chin, R., 1995. Decomposition of arbitrarily shaped morphological structuring elements. IEEE Transactions on Pattern Analysis and Machine Intelligence 17, 2–15. Reshak, A.H., 2009. Enhancing the resolution of the forward second harmonic imaging using the two-photon laser scanning microscope. Micron 40, 750–755. Serra, J., 1982. Image Analysis and Mathematical Morphology. Academic Press, New York. Serra, J., 1988. Toggle Mappings. Technical Report N-18/88/MM, Centre de Morphologie Mathematique, ENSMP, Fontainebleau, France. Schilders, S.P., Gan, X.S., Gu, M., 1998. Image enhancement in a reflection optical microscope by suppression of diffusing photons using polarizing annular objectives. Microscopy and Microanalysis 4, 415–419. Schultze-Lam, S., Harauz, G., Beveridge, T.J., 1992. Participation of a cyanobacterial S layer in fine-grain mineral formation. Journal of Bacteriology 174, 7971–7981. Shah, S., 2007. Image enhancement for increased dot-counting efficiency in FISH. Journal of Microscopy 228, 211–226. Soille, P., 2003. Morphological Image Analysis – Principle and Applications. Springer, Germany. Tang, B., Sapiro, G., Caselles, V., 2001. Color image enhancement via chromaticity diffusion. IEEE Transactions on Image Processing 10, 701–707. Wan, Y., Shi, D., 2007. Joint exact histogram specification and image enhancement through the wavelet transform. IEEE Transactions on Image Processing 16, 2245–2250. Yang, C., Lu, L., Lin, H., Guan, R., Shi, X., Liang, Y., 2008. A fuzzy-statisticsbased principal component analysis (FS-PCA) method for multispectral image enhancement and display. IEEE Transactions on Geoscience and Remote Sensing 46, 3937–3947.