Sea Cucumber Image Dehazing Method by Fusion of Retinex and Dark Channel

Proceedings, 6th IFAC Conference on Bio-Robotics Proceedings, 6th IFAC Conference on Bio-Robotics Beijing, China,6th July 13-15, 2018 Proceedings, 6th IFAC Conference on Bio-Robotics online at www.sciencedirect.com Proceedings, IFAC Conference onAvailable Bio-Robotics Beijing, China,6th July 13-15, 2018 Proceedings, IFAC Conference on Bio-Robotics Beijing, Beijing, China, China, July July 13-15, 13-15, 2018 2018 Beijing, China,6th July 13-15, 2018 Proceedings, IFAC Conference on Bio-Robotics Beijing, China, July 13-15, 2018

ScienceDirect

IFAC PapersOnLine 51-17 (2018) 796–801

Sea Cucumber Image Dehazing Method by Fusion of Retinex and Sea Dehazing Method by Fusion of Retinex and Dark Dark Channel Channel Sea Cucumber Cucumber Image Image Dehazing Method by Fusion ofFang Retinex Zhenbo Li**, ***, Guangyao Li *, Bingshan Niu*, Peng*  and Dark Channel Sea Cucumber Image Dehazing Method by Fusion of Retinex Zhenbo Li**, ***, Guangyao Li *, Bingshan Niu*, Fang Peng*  and Dark Channel Zhenbo Li**, Li**, ***, ***, Guangyao Guangyao Li Li *, *, Bingshan Bingshan Niu*, Niu*, Fang Peng* Zhenbo Peng* * Information and Electrical Engineering, China Agricultural University, Beijing, China Sea Cucumber Dehazing Method Fusion ofFang Retinex Dark Channel Zhenbo Li**, Guangyao Li *,by Bingshan Niu*, Fang Peng* and * College College of of Image Information and***, Electrical Engineering, China Agricultural University, Beijing, China

* College of Information and Electrical Engineering, China Agricultural Beijing, China (Tel:18810990726; e-mail: [email protected]). * and Electrical Engineering, China University, Beijing, Zhenbo Li**, Guangyao Li *, Bingshan Niu*, Fang University, Peng* (Tel:18810990726; e-mail: [email protected]). * College College of of Information Information and***, Electrical Engineering, China Agricultural Agricultural University, Beijing, China China (Tel:18810990726; e-mail: [email protected]). ** Key Laboratory of Agricultural Information Acquisition ,, Department of Agriculture, Beijing, China (e-mail: (Tel:18810990726; e-mail: [email protected]). [email protected]). ** Key Laboratory of Agricultural Information Acquisition Department of Agriculture, Beijing, China (e-mail: e-mail: (Tel:18810990726; * College of Information and Electrical Engineering, China Agricultural University, Beijing, China ** Key Laboratory of Agricultural Information Acquisition ,, Department of Agriculture, Beijing, China (e-mail: [email protected]) ** Key Laboratory of Agricultural Information Acquisition Department of Agriculture, Beijing, China (e-mail: ** Key Laboratory of Agricultural [email protected]) Acquisition , Department of Agriculture, Beijing, China (e-mail: (Tel:18810990726; e-mail: [email protected]). [email protected]) *** Agricultural Internet of Things Engineering Technology Research Center, Beijing, China (e-mail: [email protected])} [email protected]) *** Agricultural Internet of Things Engineering Technology Research Center, Beijing, China (e-mail: [email protected])} [email protected]) ** Key Laboratory of Agricultural Information Acquisition , Department of Agriculture, Beijing, China (e-mail: *** Agricultural Internet of Things Engineering Technology Research *** Agricultural Agricultural Internet Internet of of Things Things Engineering Engineering Technology Technology Research Research Center, Center, Beijing, Beijing, China China (e-mail: (e-mail: [email protected])} [email protected])} *** [email protected])Center, Beijing, China (e-mail: [email protected])} Abstract: This proposes aa method based on theResearch prior fusion of Retinex dark [email protected])} to enhance *** Agricultural Internet of Things Engineering Technology Center, Beijing,and China (e-mail: Abstract: This paper paper proposes based on prior of and dark to enhance enhance Abstract: This paper proposessea a method method based on the theFirstly, prior fusion fusion of Retinex Retinex and dark ischannel channel to the defogging of underwater cucumber images. the original RGB image pre-processed by Abstract: This paper proposes a method based on the prior fusion of Retinex and dark channel to enhance the defogging of underwater sea cucumber images. Firstly, the original RGB image is pre-processed by Abstract: This paper proposes a method based on the prior fusion of Retinex and dark channel to enhance the defogging of underwater sea cucumber images. Firstly, the original RGB image is pre-processed by dark channel prior, and then the reflection property of the image is preserved by weighted average of pixels the defogging of underwater sea cucumber images. Firstly, the original RGB image is pre-processed by dark channel prior, and then the reflection property of the image is preserved by weighted average of pixels the defogging of underwater sea cucumber images. Firstly, the original RGB image is pre-processed by Abstract: This paper proposes a method based on the prior fusion of Retinex and dark channel to enhance dark channel prior, and then the image reflection property of of theaimage image is preserved preserved by weighted average of pixels pixels in the image. Then, the original is convolved with Gaussian template to generate a high-frequency dark channel prior, and then the reflection property the is by weighted average of in image. Then, the original is convolved with Gaussian template to generate aaaverage high-frequency dark channel prior, and the reflection property theaaimage preserved by weighted of pixels the defogging of underwater seaimage cucumber images.of Firstly, theisoriginal RGB is pre-processed by in the the image. Then, the then original image is and convolved with Gaussian template toimage generate high-frequency enhanced image. Finally, the brightness saturation of image are enhanced by changing the values in the image. Then, the original image is convolved with aa the Gaussian template to generate aa high-frequency enhanced image. Finally, the brightness and saturation of the image are enhanced by changing the in the image. Then, the original image is convolved with Gaussian template to generate high-frequency dark channel prior, and then the reflection property of the image is preserved by weighted average of values pixels enhanced image. Finally, the brightness and saturation of the image are enhanced by changing the values of S and V in the HSV image. The experimental results are represented by four evaluation indicators such enhanced image. Finally, the and saturation of the image are enhanced by the of S and V in the HSV The experimental results are represented four indicators such enhanced image. Finally, the brightness brightness saturation image template areby enhanced by changing changing the values values in the image. Then, the image. original image is and convolved withof a the Gaussian to evaluation generate a high-frequency The are represented by four evaluation indicators such of S and V in the HSV image. as the MSE. By processing images ofexperimental sea cucumber,results we obtained MSE, ENL, EI, and SNR values of 1.9782, of S and V in the HSV image. The experimental results are represented by four evaluation indicators such as the MSE. By processing images ofexperimental sea cucumber , we obtained MSE, ENL, EI, and SNR values of 1.9782, of S and V in the HSV image. The results are represented by four evaluation indicators such enhanced image. Finally, the brightness and saturation of the image are enhanced by changing the values By processing images of sea cucumber , we obtained MSE, ENL, EI, and SNR values of 1.9782, as the MSE. 14.4049, 6.9586, and 14.9172, Compared with other methods, the image processed by our as the MSE. By processing imagesrespectively. of sea cucumber ,, we MSE, ENL, EI, and SNR of 14.4049, and 14.9172, respectively. Compared with other image processed by our as theand MSE. images ofexperimental sea cucumber we obtained obtained MSE,methods, ENL, EI,the and SNR values values of 1.9782, 1.9782, of S V6.9586, inBy theprocessing HSV image. The results are represented by four evaluation indicators 14.4049, 6.9586, and 14.9172, respectively. Compared with other methods, the image processed bysuch our method has better performance in evaluating indicators. It shows that our method shows great performance 14.4049, 6.9586, and 14.9172, respectively. Compared with other methods, the image processed by our method has better performance in evaluating indicators. It shows that our method shows great performance 14.4049, 6.9586, and 14.9172, respectively. Compared with other methods, the image processed by our as the MSE. By processing images of sea cucumber , we obtained MSE, ENL, EI, and SNR values of 1.9782, It shows that our method shows great performance performance method has better betterand performance in evaluating evaluating indicators. in the defogging enhancement of underwater sea cucumber images. method has performance in indicators. It shows that our method shows great in enhancement of sea images. method has betterand performance evaluating indicators. Itwith shows thatmethods, our method performance 14.4049, 6.9586, and 14.9172, in respectively. Compared other the shows imagegreat processed by our in the the defogging defogging and enhancement of underwater underwater sea cucumber cucumber images. in the and enhancement of underwater sea images. Keywords: channel; Retinex; HSV; Image dehazing; Image enhancement in the defogging defogging enhancement of underwater sea cucumber cucumber images. method hasDark betterand performance in evaluating indicators. It shows that our method shows great performance Keywords: Dark channel; Retinex; HSV; Image dehazing; Image enhancement Keywords: Dark channel; Retinex; HSV; Image dehazing; Image enhancement Keywords: Dark channel; Retinex; HSV; Image Image enhancement © the 2018, IFAC (International Federation of Automatic Hosting by Elsevier Ltd. All rights reserved. in defogging and enhancement of underwater seaControl) cucumber images. Keywords: Dark channel; Retinex; HSV; Image dehazing; dehazing; Image enhancement    Keywords: Dark channel; Retinex; HSV; Image dehazing; Image enhancement on, without prior knowledge, it can obtain aa better quality color on, without prior knowledge, it can obtain quality color  1. INTRODUCTION on, without prior knowledge, it can obtain aa better better quality color 1. INTRODUCTION image. on, without prior knowledge, it can obtain better quality color 1. INTRODUCTION INTRODUCTION image. on, without prior knowledge, it can obtain a better quality color 1.  image. 1. INTRODUCTION Sea cucumber products contain rich protein and vitamins, and image. image. Sea cucumber products contain rich protein and vitamins, and 2. RELEATED WORK on, without prior knowledge, it can obtain a better quality color Sea cucumber products contain richfat protein and vitamins, and 2. RELEATED WORK have the characteristics of low and good balance of 1. INTRODUCTION Sea cucumber products contain rich protein and vitamins, and 2. RELEATED WORK have the characteristics of low fat and good balance of Sea cucumber products contain rich protein and vitamins, and image. 2. RELEATED WORK have the characteristics of low fat good balance of 2. RELEATED WORK enhancement has nutrition, and become an important source for people to In recent years, research based have the characteristics of low fat and and good balance of nutrition, become an source for people to In recent years, research based on on image image enhancement enhancement has has have the and characteristics ofimportant low and good balance of Sea cucumber products contain richfat protein and vitamins, and nutrition, and become an important source for people to In recent years, research based on image consume high quality animal protein (Hu et.al,2015). The focused on the performance of underwater image quality 2. RELEATED WORK nutrition, and become an important source for people to In recent years, research based on image enhancement has consume high quality animal protein (Hu et.al,2015). The focused on the performance of underwater image quality nutrition, and become an important source for people to In recent years, research based on image enhancement has have the high characteristics of low fat and good balanceThe of focused on the performance of underwater image quality consume quality animal protein (Hu et.al,2015). shape, size, color and texture of sea cucumbers play an degradation selected corresponding image enhancement consume high quality animal protein (Hu et.al,2015). The focused on and the performance of underwater image quality shape, size, and texture of sea cucumbers play an degradation and selected corresponding image enhancement consume high quality animal protein (Hu et.al,2015). The the research performance image quality nutrition, andcolor become an important source for people to focused In recenton years, basedofonunderwater image enhancement has shape, size, color and texture of sea cucumbers play an degradation and selected corresponding image enhancement important role in sea cucumber breeding. It can not only reflect techniques, such as histogram equalization (Yang et.al,2016), shape, size, color and texture of sea cucumbers play an degradation and selected corresponding image enhancement important role in sea cucumber breeding. It can not only reflect techniques, such as histogram equalization (Yang et.al,2016), shape, size, color and texture of sea cucumbers play an degradation and selected corresponding image enhancement consume high quality animal protein (Hu et.al,2015). The focused on the performance of underwater image quality important role in sea cucumber breeding. It can not only reflect techniques, such as histogram equalization (Yang et.al,2016), the basic growth of sea cucumber, but also provide reference Low-pass filter (Srividhya et.al,2015), and(Yang wavelet transform important role in cucumber breeding. It can not reflect techniques, such as histogram equalization et.al,2016), the basic growth of sea cucumber, provide reference Low-pass filter et.al,2015), and wavelet transform important rolecolor in sea sea cucumber Itcucumbers can not only only reflect such as histogram equalization (Yang et.al,2016), shape, size, and texturebreeding. ofbut seaalso play an techniques, degradation and(Srividhya selected corresponding image enhancement the basic growth of sea cucumber, but also provide reference Low-pass filter (Srividhya et.al,2015), and wavelet transform for the feeding, fishing and grading of sea cucumber, and (Singh et.al,2015). the basic growth of sea cucumber, but also provide reference Low-pass filter (Srividhya et.al,2015), and wavelet transform for the feeding, fishing and grading cucumber, and (Singh et.al,2015). et.al,2015). the basic growth of sea cucumber, but of also provide reference Low-pass filter et.al,2015), and(Yang wavelet transform important role in sea cucumber breeding. Itsea can not only reflect techniques, such(Srividhya as histogram equalization et.al,2016), for the feeding, fishing and grading of sea cucumber, and (Singh provide data support for the monitoring of breeding for the feeding, fishing and grading of sea cucumber, and (Singh et.al,2015). provide data support for the monitoring of breeding for the feeding, fishing and grading of sea cucumber, and (Singh et.al,2015). Hitam (Hitam et al., 2013) proposed a hybrid contrastthe basic growth of sea cucumber, but also provide reference Low-pass filter (Srividhya et.al,2015), and wavelet transform provide data support for the monitoring of breeding Hitam (Hitam et al., 2013) proposed aa hybrid contrastenvironment (Zhu et.al,2008). At present, the use of computer provide data support for the monitoring of breeding Hitam (Hitam et al., 2013) proposed hybrid contrastenvironment (Zhu et.al,2008). At present, the use of computer provide data support for the monitoring of breeding constrained histogram equalization method for underwater for the feeding, fishing and grading of sea cucumber, and (Singh et.al,2015). Hitam (Hitam et al., 2013) proposed a hybrid contrastenvironment (Zhu et.al,2008). At present, the use ofacomputer constrained histogram method for underwater Hitam (Hitam et al., equalization 2013) proposed a hybrid contrastvision technology estimate biomass has become research environment (Zhu to et.al,2008). At the use constrained histogram equalization method for vision technology to estimate has become research environment et.al,2008). At present, present, the useofof ofaacomputer computer image enhancement. enhancement. Aiming at the the problem problem that theunderwater quality of of provide data(Zhu support for biomass the monitoring breeding constrained histogram equalization method for underwater vision technology to estimate biomass has become research image Aiming at that the quality constrained histogram equalization method for underwater hotspot[ ]. It not only effectively reduces human and material vision technology to estimate biomass has become a research Hitam (Hitam et al., 2013) proposed a hybrid contrastimage enhancement. Aiming at the problem that the quality of hotspot[ ]. It not only effectively reduces human and material vision technology to estimate biomass has become a research underwater image is greatly affected by illumination and environment (Zhu et.al,2008). At present, the use of computer image enhancement. Aiming at the problem that the quality of hotspot[ ]. It Itbut notalso only effectively reduces human and material underwater image is greatly by illumination image enhancement. Aiming at affected the problem that qualityand of resources, quickly and accurately obtain various constrained histogram equalization method fortheunderwater hotspot[ ]. not only effectively reduces human and material underwater image is greatly by illumination resources, quickly and accurately obtain various hotspot[ ]. Itbut notalso only effectively reduces and material turbidity, Qiao Qiao Xi (Qiao (Qiao et al., al.,affected 2017) adopts adopts the method methodand of vision technology to estimate biomass hashuman become a research underwater image is greatly affected by illumination and resources, but also quickly and accurately obtain various turbidity, Xi et 2017) the underwater image is greatly affected by illumination and statistical information of sea cucumber, providing a basis for resources, but also quickly and accurately obtain various image enhancement. Aiming at the problem that the quality of turbidity, Qiao Xi (Qiao (Qiao et al., al., histogram 2017) adopts adopts the method methodand of statistical information of sea cucumber, providing a basis for resources, but also quickly and accurately obtain various constrained contrast adaptive equalization hotspot[ ]. It not only effectively reduces human and material turbidity, Qiao Xi et 2017) the of statistical information of sea cucumber, providing aa basis for constrained contrast histogram turbidity, Qiao Xi (Qiao et al.,affected 2017) adopts the methodand of efficient and economical breeding of sea cucumber. statistical information of sea providing for image is adaptive greatly by equalization illumination constrained contrast adaptive histogram equalization and efficient and economical sea cucumber. statistical information of breeding sea cucumber, cucumber, providing a basis basis for underwater resources, but also quickly and of accurately obtain various wavelet transform transform to adaptive enhance histogram the image. image. equalization The underwater underwater constrained contrast and efficient and economical breeding of sea cucumber. wavelet to enhance the The constrained contrast adaptive histogram equalization and turbidity, Qiao Xi (Qiao et al., 2017) adopts the method of efficient and economical breeding of sea cucumber. wavelet transform to enhance enhance thenon-uniform image. The Theillumination, underwater efficient and economical breeding of sea providing cucumber. Due to the scattering, absorption effect of the light and the image noise noise will increase increase due to to statistical information of sea cucumber, a basis for wavelet transform to the image. underwater Due to the scattering, absorption effect of the light and the image will due non-uniform illumination, wavelet transform to enhance the image. The underwater constrained contrast adaptive histogram equalization and Due to to the the absorption effect of the the the lightcollected and the the image noise will increase(Emberton due to to non-uniform non-uniform illumination, interference of underwater suspended microbes, etc.will Emberton et al., al., 2015) 2015) The image efficient andscattering, economical breeding of seamatter, cucumber. Due absorption effect of light and noise increase due illumination, interference of underwater suspended microbes, etc. Emberton (Emberton et The image Due to the scattering, scattering, absorption effectmatter, of the the lightcollected and the image image will increase due to waveletnoise transform to enhance thenon-uniform image. Theillumination, underwater interference of underwater suspended matter, the collected microbes, etc. Emberton (Emberton et al., 2015) The image images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, interference of underwater suspended matter, the collected microbes, etc. Emberton (Emberton et al., 2015) The image images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, interference of underwater suspended matter, the collected microbes, etc. Emberton (Emberton et al., 2015) The image Due to the scattering, absorption effect of the light and the image noise will increase due to non-uniform illumination, images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, interference of underwater suspended matter, the collected microbes, etc. Emberton (Emberton et al., 2015) The image distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored images of the sea cucumber have serious fogging and color layering method is used to estimate the atmospheric light, optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven distortion, which restricts the development of the underwater which avoids the oversaturation phenomenon of the restored cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have quality. Traditional methods have contrast enhancement achieved certain effects, because of the lack of consideration cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have quality. Traditional methods have contrast enhancement achieved certain effects, because of the lack of consideration cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have optical vision technology. Therefore, the underwater sea image, but is easily affected by the image noise and uneven quality. Traditional Traditional methods histogram have contrast contrast enhancement achieved certain effects, effects, because of the theimage lack of of consideration techniques (Wang et.al,1999), equalization (Chen of the underlying causes of underwater degradation, the quality. methods have enhancement achieved certain because of lack consideration techniques (Wang et.al,1999), histogram equalization (Chen of the underlying causes of underwater image degradation, the quality. Traditional methods have contrast enhancement achieved certain effects, because of the lack of consideration cucumber images need to be enhanced to enhance the visual light conditions. These image enhancement methods have techniques (Wang (Wang et.al,1999),these histogram equalization (Chen of the underlying underlying causes of underwater image degradation, the et.al,2012), etc. However, methods often lead to enhanced results often fail to accurately reflect the actual techniques et.al,1999), histogram equalization (Chen of the causes of underwater image degradation, the et.al,2012), etc. However, these methods often lead to enhanced results often fail to accurately reflect the actual techniques (Wang et.al,1999), histogram equalization (Chen of the underlying causes of underwater image degradation, the quality. Traditional methodsthese havemethods contrast often enhancement certain effects, because of the lackreflect of consideration et.al,2012), etc.enhancement However, lead to to achieved enhanced results often fail fail to accurately accurately the actual actual excessive image and whitening. appearance of the images. et.al,2012), etc. However, these methods often lead enhanced results often to reflect the excessive image enhancement and whitening. appearance of the images. et.al,2012), etc. However, these methods often lead to enhanced results often fail to accurately reflect the actual techniquesimage (Wang et.al,1999),and histogram equalization (Chen appearance of the underlying excessive enhancement whitening. of thecauses images.of underwater image degradation, the excessive image enhancement and appearance of excessive image enhancement and whitening. whitening. appearance of the the images. images. In this paper, a comprehensive method based on the et.al,2012), etc. However, these methods often lead to enhanced results often fail to accurately reflect the actual For this purpose, by analyzing analyzing the imaging mechanism of In this paper, aa comprehensive method based on the For this purpose, by the imaging mechanism of In this paper, comprehensive method based on the For this purpose, by analyzing the imaging mechanism of transcendental fusion of Retinex and dark channel is proposed, underwater images, Roser (Roser et al., 2014) established an excessive image enhancement and whitening. appearance of the images. In this paper, a comprehensive method based on the For this purpose, by analyzing the imaging mechanism of transcendental fusion of Retinex and dark channel is proposed, underwater images, Roser (Roser et al., 2014) established an In this paper, a comprehensive method based on the For this purpose, by analyzing the imaging mechanism of transcendental fusion of Retinex and dark channel is proposed, underwater images, Roser (Roser et al., 2014) established an which combines HSV space color enhancement theory to imaging model for enhancing image sharpness transcendental fusion of Retinex and dark channel is proposed, underwater images, Roser (Roser et al., 2014) established an which combines HSV space color enhancement theory to imaging model for enhancing image sharpness transcendental fusion of Retinex and dark channel is proposed, underwater images, Roser (Roser et al., 2014) established an In this paper, a comprehensive method based on the For this purpose, by analyzing the imaging mechanism of which combines HSV space color enhancement theory to underwater imaging for enhancing image sharpness avoid color distortion while maintaining the freshness of the under natural natural light model and turbid turbid conditions. Seemakurthy which combines HSV space color enhancement theory to imaging model for image sharpness avoid color distortion the freshness of the under light and Seemakurthy which combines HSV spacemaintaining color enhancement theory to underwater imaging model for enhancing enhancing image sharpness transcendental fusion ofwhile Retinex and dark channel is proposed, underwater images, Roser (Roser etconditions. al., 2014) established an avoid color distortion while maintaining the freshness of the under natural light and turbid conditions. Seemakurthy image. The experimental show that this method is (Seemakurthy et.al, 2015) proposes mathematical model avoid color distortion whileresults maintaining the of under natural light and conditions. Seemakurthy image. The experimental results show that this method is (Seemakurthy et.al, 2015) proposes aaa mathematical model avoid color distortion maintaining the freshness freshness of the the under natural light model and turbid turbid conditions. Seemakurthy which combines HSVwhile space color enhancement theory to underwater imaging for enhancing image sharpness image. The experimental results show that this method is (Seemakurthy et.al, 2015) proposes mathematical model simple and can solve the problem of pseudo shadow for the motion blurred image captured by light refraction due image. The experimental results show that this method is (Seemakurthy et.al, 2015) proposes a mathematical model simple and can solve the problem shadow for the motion image captured light refraction due image. The experimental results show of that this method is (Seemakurthy et.al, proposes aby mathematical model avoid color distortion while maintaining the pseudo freshness of the under natural blurred light 2015) and turbid conditions. Seemakurthy simple and can solve the problem of pseudo shadow for the motion blurred image captured by light refraction due phenomenon and color distortion, noise amplification and so to the refraction of light. The unidirectional cyclic wave is simple and can solve the problem of pseudo shadow for the motion blurred image captured by light refraction due phenomenon and color distortion, noise amplification and so to the refraction of light. The unidirectional cyclic wave is simple and can solve the problem of pseudo shadow for the motion blurred image captured by light refraction due image. The experimental results show that this method is (Seemakurthy et.al, 2015) proposes a mathematical model phenomenon and color distortion, noise amplification and so to the refraction of light. The unidirectional cyclic wave is regarded as a spatial invariant for underwater blurred image phenomenon and color distortion, noise amplification and so to the refraction of light. The unidirectional cyclic wave is regarded as a spatial invariant for underwater blurred image phenomenon noise amplification and so for to the refraction of light. The unidirectional cyclic wave is simple and and can color solvedistortion, the problem of pseudo shadow the motion blurred image captured by light refraction due regarded as aa spatial invariant for underwater blurred image regarded as spatial invariant for underwater blurred image regarded a spatial invariant for underwater cyclic blurredwave image 2405-8963 © 2018, of Automatic Control) by as Elsevier Ltd. All rights phenomenon andIFAC color(International distortion, Federation noise amplification and so Hosting to the refraction of light. Thereserved. unidirectional is Copyright 2018 responsibility IFAC 797Control. Peer review© of International Federation of Automatic Copyright ©under 2018 IFAC 797 regarded as a spatial invariant for underwater blurred image Copyright © 10.1016/j.ifacol.2018.08.098 Copyright © 2018 2018 IFAC IFAC Copyright © 2018 IFAC

797 797 797

IFAC BIOROBOTICS 2018 Beijing, China, July 13-15, 2018

Zhenbo Li et al. / IFAC PapersOnLine 51-17 (2018) 796–801

recovery. The single method is not ideal for the treatment of underwater images. Image fusion is gradually applied to underwater image enhancement (Ancuti et al., 2012). Color compensation, histogram equalization and other enhancement techniques can be used to process underwater images. Celebi (Celebi et.al, 2012) proposed the use of empirical mode decomposition for underwater image enhancement processing, through the original image signal is decomposed into a series of intrinsic mode function, then use genetic algorithm to calculate each modal function to get the weights of spectral channels component, to improve the image visual quality, the method than the traditional contrast stretching and histogram equalization can get better results.

797

There are three main factors that result in low channel values in dark primary colors: a) projection of natural scenery; b) colorful objects or surfaces in some of the three channels of RGB have very low values; c) darker colors Object or surface. There are shadows or colors everywhere in the images of underwater sea cucumbers. The fog map formation model normally described by the following equation:

I  x   J  x  t  x   A 1  t  x   Where

I  x  is the image to be defogged, J  x  is the

fog-free image we want to recover,

With the development of deep learning techniques, ****，such as Yang (Yang et al., 2017) proposed a joint optimization based on image content and texture constraint of multi-scale neural plaques synthetic method to achieve the latest repair precision. Then because of the underwater acquisition image atomization phenomenon is more serious, so the method based on deep learning remains to be breakthrough.

t  x  is the transmittance.

The current known condition is

I  x  , which requires the

J  x  . Through the above formula, our goal is

to calculate the original fog-free image, the transmittance, and the estimated global atmospheric light composition from the existing photographs.

The contributions of this paper are:(1) Increase the exposure between the Retinex and dark channel prior methods to increase the brightness of the pre-processed image. (2) Set the range of defogging coefficient  according to the color characteristics of sea cucumber. (3) Combining HSV spatial color theory to enhance sea cucumber image saturation.

Fig. 1. Underwater sea cucumber image fog principle

Fig.1 This image describes the principle of sea cucumber imaging, which is an image transmitted directly and partially scattered, and a final image formed by mixing with global

3.IMAGE ENHANCEMENT MODEL 3.1 Dark channel prior algorithm

atmospheric light components, that is The de-fog algorithm based on the dark channel prior is a statistical algorithm. In most regions, some pixels always have at least one color channel with a very low value. That is, the minimum value of the light intensity in this area is a very small number. For any input image J , the dark channel can be expressed by the following expression:

Where J

c

t  x   ex Where

(4)

is the parameter for atmospheric scattering and

The basic flow of the defogging algorithm based on the dark channel prior is as follows:

calculate the minimum value of the RGB component of each pixel, store it in a grayscale image with the same size as the original image, and then perform the minimum filtering on this grayscale image. The filter radius is determined by the size of the window. The theory of dark channel priors indicates:

 0.



  d  x

d is the depth of field. This formula expresses that scene radiance exhibits exponential attenuation with increasing distance. Because if we get this transmittance, we can get the depth of the scene according to this law.

(1)

  x  denotes a window centered on pixel x . First

J

I  x .

For the calculation of the transmittance, there is an existing formula that indicates that when the material in the atmosphere is homogeneous, the transmittance t can be expressed as:

denotes each channel of a color image, and

dark

A is the global

atmospheric light component, and

target value

Since the zooplankton in the aquaculture water in the offshore area has increased the difficulty of image defogging and enhancement, we proposed a method based on the prior fusion of Retinex and dark channel for image enhancement and recovery of underwater sea cucumber based on the characteristics of near-seawater breeding water quality. It has improved the quality of underwater sea cucumber images.

J dark  x   min  min J c  y   y x   cr , g ,b 

(3)

(1) The transmittance is obtained. According to formula (3), the equation can be divided by both sides and then converted.

  I c  y  t  x   1   min  min    c y x   c  A  

(2)

798

(5)

IFAC BIOROBOTICS 2018 798 Beijing, China, July 13-15, 2018

Zhenbo Li et al. / IFAC PapersOnLine 51-17 (2018) 796–801

  I c  y  In fact, min  min    is the dark channel of the c y x   c  A   I c  y regularized image . Considering the existence of the Ac

(4) Opposite

R  x, y   exp G  x, y  RHSV  x, y  

N G  n log I s  x, y   log  Fn  x, y   I s  x, y    n1 



Where

max  t  x  , T0 

weight corresponding to the dimension



n

 1;

4. EXPERIMENT AND ANALYSIS 4.1 Simulations Simulations were performed on MatLab R2017a to verify the enhancement of the underwater sea cucumber image using the Retinex and dark channel prior fusion method. First, the gray scale image of the underwater sea cucumber was selected as the object of the simulation and the dark channel prior fusion was performed. Retinex processing and gradation transformation for image enhancement processing, select a few typical images in the experiment (if there is a single sea cucumber, more than sea cucumber, with or without attachment base, etc.), the effect of the comparison chart shown in Fig. 3 to Fig. 5. From Fig. 3 to Fig. 5, we can see that among the seven methods used in the experiment, our method gets the best effect, and the multi-scale Retinex method is better, while the other four methods are obviously worse from the visual sense. It is based on the human subject's subjective visual system to establish a mathematical model, and through specific formulas to calculate the quality of the image. In this paper, the mean square error (MSE), equivalent coefficient (NEL), information entropy (IE), and signal-to-noise ratio (SNR) are calculated between the processed image and the original image. Evaluation indicators to compare effectiveness. The MSE reflect the mathematical statistical differences between the reconstructed image and the original image.

(1) Logarithmically separates the irradiated light component

L  x, y  and the reflected light component R  x, y  .

(7)

(2) The reflection image is hypothetically estimated as a spatially smooth image, and the original image is convoluted with a Gaussian template to obtain a low-pass filtered image

D  x, y  , F  x, y  denotes a Gaussian filter function.

(8)

4.2 Evaluation System

(3) In the logarithmic domain, the low-pass filtered image is subtracted from the original image to obtain a high-frequency

(1) Mean Square Error (MSE). MSE represents the mean square error of the current image X and the reference image Y , and H , W are the height and width of the image, respectively. In general, the smaller the MSE after image processing, the better the processing effect.

G  x, y  .

G  x, y   logS  x, y   log D  x, y 

n and

G represents the gain coefficient, which is taken as 1 in the experiment.

(6)

The priori of dark colors is a focal result and is the result of a large number of outdoor fog-free photos. If the target scene is intrinsically similar to atmospheric, such as snow, white walls, and the sea, the prerequisites are Not. Correctly, it can’t obtain fairly results. Therefore, we use the Retinex and dark channel prior fusion, combined with HSV theory to enhance the effect of image defogging. The basic steps of the improved dark channel prior algorithm are as follows.

enhanced image

is the

n 1

3.2 Improved dark channel prior algorithm

D  x, y   logS  x, y   F  x, y 

n N

(3)About the setting of  . Parameter  is an adjustment parameter. When this parameter is 0, it means that the fog is not removed. If the parameter is 1, it means that all the fog is removed.

log J  x, y   log R  x, y   log L  x, y 

N is the number of scales. I s  x, y  is the image

after the RGB image is converted to the HSV.

t  T0 . Therefore, the final recovery formula is as follows: A

   

(11)

(2) Get a restored image. When the value of the projective graph t is small, the value of J will be too large, so that the whole image will be over-whitened. Therefore, We set a threshold T0 . When the value of t is less than T0 , let

I  x  A

(10)

(5) Enhanced based on HSV theory.

atmospheric perspective phenomenon, if the fog of the image is completely removed, it will look unnatural and may lose the depth information. Therefore, it is necessary to keep a small amount of fog on the distant objects, so it is introduced in the formula. The parameter  , we set value is 0.90.

J  x 

G  x, y  to get an enhanced image R  x, y  .

(9)

799

IFAC BIOROBOTICS 2018 Beijing, China, July 13-15, 2018

Zhenbo Li et al. / IFAC PapersOnLine 51-17 (2018) 796–801

799

Table 1 Quantitative results in terms of MSE, ENL,IE and SNR for images in Figs.3-Figs.5 Approach Our method[] Gaussian Dark Channel Prior Adaptive Histogram Equalization Adaptive color scale and contrast Histogram equalization Multiscale Retinex

MSE 1.9782 9.8848 6.1116 3.5370 30.4686 34.4957 2.2544



ENL 14.4049 43.1130 45.9172 58.3764 71.0088 3.0148 27.1569

EI 6.9586 6.8400 6.2426 5.8557 6.8741 6.6461 6.4470

SNR -14.9172 -9.0768 -11.6342 -3.4035 -6.4212 -16.6520 -14.2921

Figure 2 Evaluation Index Histogram

(2) Equivalent numbers of looks(ENL). This is an indicator of the smoothness of the uniform area. In general, we can select several regions of interest to find the equivalent coefficients and then average them, or directly determine the equivalent visual number of the image. 2  1  H h ENL     2  H  h 1  h

(13)

Where H denotes the number of selected regions,  denotes the average of selected region pixels, and  denotes the mean squared error of the selected region (3) Information Entropy(IE). Entropy is a measure of the diversity or uniformity of microscopic states in thermodynamics and represents the disorder of the system. Assume that a system X may be in a different centralized state

Fig. 3. Single sea cucumber image enhancement process diagram

MSE 

1 H W

H

W

  X i, j   Y i, j  

2

x1 , x2 ,

, xn , p  x  represents the probability of

occurrence of a state information entropy

(12)

follows.

i 1 j 1

Where i and j respectively represent the horizontal and vertical coordinates of the image pixels. 800

xi  i  1, 2,

, n  , then the

H  x  of the system is defined as

IFAC BIOROBOTICS 2018 800 Beijing, China, July 13-15, 2018

Zhenbo Li et al. / IFAC PapersOnLine 51-17 (2018) 796–801

Fig. 4. Different methods for image processing of single sea cucumber

Fig. 5. Different methods for image processing of multiple sea cucumbers

The collected sea cucumber images were processed, and the values of MSE, ENL, EI and SNR in Table 1 were obtained. From the table, we can observe that the image MSE based on the prior fusion method of Retinex and the dark channel has the best effect value, ENL, information entropy, SNR effect value is good, which is among 4 indicators. According to the analysis data in the table, the method based on the prior fusion of the Retinex and the dark channel has a better effect on the enhancement of underwater sea cucumber images. As shown in Fig.2, the result of each evaluation index is visually displayed as a bar graph. Since the value of the SNR indicator is negative, the value of the SNR is represented as a reciprocal number.

n

H  x    p  x  log  p  x   i 1

Where

(14)

0  p  x   1 and

n

 p  x =1 .

The greater the

i 1

information entropy, the higher the disorder of the information and the greater the amount of information contained. (4) Signal to Noise Ratio (SNR). If we regard the sum of the original graph

f  x, y  as

g  x, y  and the noise image

e  x, y  , the mean square noise of the output graph is as

5. CONCLUSIONS

follows. M 1 N 1

M 1 N 1

 f  x, y  /   f  x, y   g  x, y  x 0 y 0

2

x 0 y 0

This article takes the underwater sea cucumber image as the research object, aiming at the phenomenon of blurred image. Based on the prior fusion method of Retinex and dark channel, combined with the theory of HSV spatial color enhancement, the phenomenon of color distortion is avoided while maintaining the vividness of the image. The experimental results show that this method is simple and can solve problems such as color distortion and serious fogging, and can obtain better quality color images without prior knowledge. By evaluating the four methods of comparison and fusion of the function MSE, ENL, IE and SNR, we can know that our algorithm shows better performance in the defogging and

2

(15)

The Table 1 shows that the above five images in Fig. 3 to Fig. 5 are based on Retinex and dark channel prior fusion processing, combined with HSV enhancement theory transformed MSE values, ENL values, IE values and SNR values. The results show that our fusion algorithm has better performance in image defogging enhancement.

801

IFAC BIOROBOTICS 2018 Beijing, China, July 13-15, 2018

Zhenbo Li et al. / IFAC PapersOnLine 51-17 (2018) 796–801

enhancement of underwater sea cucumber images. However, the disadvantage of this method is the increased processing time. In the subsequent experiments, the model will continue to be optimized and the time efficiency will be improved. 6. ACKNOWLEDGEMENTS This research was supported by International Science & Technology Cooperation Program of China (No.2015DFA00530), Key Research and Development Research Project of Shandong Province (No.2016CYJS03A02) and National Nature Science Foundation of China Project (No.61471133，61472172). REFERENCES A. T. Çelebi, & S. Ertürk. (2012). Visual enhancement of underwater images using Empirical Mode Decomposition and wavelet denoising. Signal Processing and Communications Applications Conference,39,1-4. Bekaert, P., Haber, T., Ancuti, C. O., & Ancuti, C. (2012). Enhancing underwater images and videos by fusion. Computer Vision and Pattern Recognition,6428,81-88. Chen, S. D. (2012). A new image quality measure for assessment of histogram equalization-based contrast enhancement techniques. Digital Signal Processing, 22(4), 640-647. Emberton, S., Chittka, L., & Cavallaro, A. (2015). Hierarchical rank-based veiling light estimation for underwater dehazing. British Machine Vision Conference,125,1-12. Hitam, M. S., Awalludin, E. A., Wan, N. J. H. W. Y., & Bachok, Z. (2013). Mixture contrast limited adaptive histogram equalization for underwater image enhancement. International Conference on Computer Applications Technology, 1-5. Hu, J., Wang, J., Zhang, X., Fu, Z., & University, C. A. (2015). Research status and development trends of information technologies in aquacultures. Transactions of the Chinese Society for Agricultural Machinery, 46(7), 251-263. Li, J., Skinner, K. A., Eustice, R. M., & Johnson-Roberson, M. (2018). Watergan: unsupervised generative network to enable real-time color correction of monocular underwater images. IEEE Robotics and Automation Letters, 3(1),387-394. Qiao, X., Bao, J., Zeng, L., Zou, J., & Li, D. (2017). An automatic active contour method for sea cucumber segmentation in natural underwater environments. Computers & Electronics in Agriculture, 135(C), 134-142. Roser, M., Dunbabin, M., & Geiger, A. (2014). Simultaneous underwater visibility assessment, enhancement and improved stereo. IEEE International Conference on Robotics and Automation, 3840-3847. Seemakurthy, K., & Rajagopalan, A. N. (2015). Deskewing of underwater images. IEEE Trans Image Process, 24(3), 1046. Singh, G., Jaggi, N., Vasamsetti, S., Sardana, H. K., Kumar, S., & Mittal, N. (2015). Underwater image/video enhancement using wavelet based color correction (WBCC) method. Underwater Technology, 1-5. Srividhya, K., & Ramya, M. M. (2015). Performance analysis of pre-processing filters for underwater images.

802

801

International Conference on Robotics, Automation, Control and Embedded Systems, 1-7. Wang, Y., Chen, Q., & Zhang, B. (1999). Image enhancement based on equal area dualistic sub-image histogram equalization method. IEEE Transactions on Consumer Electronics, 45(1), 68-75. Yang, C., Lu, X., Lin, Z., Shechtman, E., Wang, O., & Li, H. (2016). High-resolution image inpainting using multiscale neural patch synthesis. Yang, W., Xu, Y., Qiao, X., Rao, W., Li, D., & Li, Z. (2016). Method for image intensification of underwater sea cucumber based on contrast-limited adaptive histogram equalization. Transactions of the Chinese Society of Agricultural Engineering, 32(6):197-203. Zhu, C. R. (2008). Application of computer vision technology in aquaculture. Journal of Zhejiang Ocean University, 27(4),439-443.