- Email: [email protected]
A new real-time processing system for the IRFPA imaging signal based on DSP&FPGA
Infrared Physics & Technology 46 (2005) 277–281 www.elsevier.com/locate/infrared
A new real-time processing system for the IRFPA imaging signal based on DSP&FPGA Hui-Xin Zhou *, Rui Lai, Shang-Qiang Liu, Bing-Jian Wang, Qing Li Lab. of Infrared Technology, School of Technical Physics, P.O. Box 204, Xidian University, Xi’an 710071, PR China Received 10 December 2003 Available online 6 July 2004
Abstract The principle, configuration and special features of a new high-precision real-time signal processing system for Infrared Focal Plane Arrays is presented in this note. A structure based on DSP&FPGA is adopted in the system. The FPGA implements the system timing control and the low level algorithms. The DSP performs the high-level algorithms, such as nonuniformity correction for the IRFPA. Nonuniformity correction and image enhancement and display algorithm, the two crucial algorithms for IR imaging signal processing, are also discussed in this note. The experiments on real IR imaging sequences demonstrate that the system is suitable for the real-time high-speed IR imaging system with high quality and high precision. 2004 Elsevier B.V. All rights reserved. PACS: 07.57.)c; 07.57.kp; 42.30.va Keywords: Infrared Focal Plan Arrays; Infrared imaging; Nonuniformity correction; Image processing; DSP; FPGA
1. Introduction With the development of Infrared focal plane array (IRFPA) technology, the advantages of high density, excellent performance, high reliability and miniaturization are available in Infrared (IR) imaging systems [1]. At present, acquisition of high quality images has become the key problem of IR imaging systems. Such systems generally need to process mass data in real-time [2]. The processing
*
Corresponding author. E-mail addresses: [email protected] Zhou), [email protected] (R. Lai).
(H.-X.
includes various algorithms, such as nonuniformity correction, image segmentation, local characteristics extraction, image de-noising, image enhancement, etc. hence then must be one well integrated high-speed information processing system. In this paper a new IRFPA imaging system based on the DSP&FPGA is presented to fulfill such requirements. Our note also focuses on two algorithms that are used in the IR image processing stage. One is the nonuniformity correction and the other is image enhancement and display method. Nonuniformity correction is a key problem that must be first solved in IRFPA systems. The detector-to-detector responsivity (gain) and dark current (offset)
1350-4495/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.infrared.2004.02.003
278
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
variations [3–7], which cause the nonuniformity, may completely mask the useful thermal signatures in an IR image with a fixed pattern ‘‘noise’’. In order to get better resolution, a 12 or 14 bit ADC is commonly adopted in the signal processing system [2]. How to effectively map 12 or 14 bit data to 8 bit TV video data is another important technology in IR image processing systems [8,9].
2. New real-time processing system for IRFPA imaging signal 2.1. Hardware configuration of the system The schematic diagram of the real-time signal processing system for IRFPA imaging based on DSP&FPGA is shown in Fig. 1. The main function unit of the system consists of an IRFPA driving circuit, an analog imaging signal processor, an ADC, a high-speed signal processor, a DAC and a video synthesizer, etc. The IRFPA driving circuit unit provides the pulse-driving signal and the offset voltage, which are crucial for the IRFPA operating state. It makes the IRFPA work at the optimum operating point. In the traditional method, the driving circuit would be complex and nonflexible. In our system, the driving-pulse signal is achieved by an FPGA. The method has the advantages of high reliability and flexibility. The pre-processing circuit unit amplifies the imaging analog signal with low noise to the level that the ADC requires. The ADC transforms the analog image signal to a digital one. In order to be
applied to image data processing with high speed and precision, a 12 bit ADC whose sampling frequency is up to 20 MHz is selected, so that a high resolution of digitized image data is obtained. The synchronization and timing control unit harmonizes the other units in the system, including the output circuit unit of the IRFPA, the ADC sample unit, the data store, the DSP, the DAC, the video synthesizer, etc. The DSP&FPGA unit is the kernel of the system. In the system of real-time IR signal processing, the low-level processing algorithms that only need simple computation deal with a large amount of data. But they do not need high-processing speed. The FPGA is adapted to doing these, and flexibility is also achieved. The high-level algorithms process smaller amounts of data, but are more complex. The DSP (ADSP21060) has the characteristics of high-speed operation, flexible addressing method and powerful communication capability. Thus the DSP is perfect for implementing the high-level algorithm. The memory must be expanded for storing a large amount of IRFPA imaging data. In this note, SRAM is used as expanded data memory, which has the advantages of fast access and lager capacity in a single chip. As an embedded DSP system, the off-chip program memory is also needed. FLASH memory is the preferred choice, which can do online erasing and programming. It also enhances the flexibility of the system. The image data processed by the DSP&FPGA module are sent to the DAC through a data buffer, where the digital image data are transformed back to an analog image signal. Under the Synchroni-
RESET
IRFPA
Preprocessing
ADC
FPGA
DSP
SRAM
Driving Circuit
FLASH
Data Buffer
DAC
Video Synthesizer
Synchronization and Timing Control Module
Fig. 1. Schematic diagram of the signal processing system for IRFPA imaging.
Monitor
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
zation signal, the analog image signal is synthesized to standard video signal in the Video synthesizer unit. Finally the image is displayed in real-time on the monitor.
1 2
gray level
2.2. Software for the algorithm
3
ϕ1
Using the technology of DSP&FPGA, with a real-time requirement, we can adopt different processing algorithms to different IR imaging signals. In the limited space of the note, only two crucial algorithms for IR imaging signal processing are discussed. 2.2.1. Nonuniformity correction algorithm The so-called nonuniformity of IRFPA is the variation of response output between the detectors in the IRFPA under uniform background illumination. There are several factors causing nonuniformity. The main sources of nonuniformity are: (1) responsivity nonuniformity, including spectral response nonuniformity; (2) nonuniformity of the readout circuit and the coupling between the detector and the readout circuit; and (3) nonuniformity of dark current [4]. Without nonuniformity compensation (NUC), the images from the IRFPA are distorted and are not suitable for field use. The individual detector outputs are linear and stable in time, so they can be expressed as [3–7,10] Vi;j ðuÞ ¼ Ri;j ðuÞu þ vi;j ðuÞ;
ð1Þ
where ði; jÞ are the coordinates of a detector in the array, u represents the irradiance that is incident on the detector ði; jÞ, Vi;j ðuÞ is the output of the detector ði; jÞ, and Ri;j ðuÞ and vi;j ðuÞ are the gain and the offset of the detector ði; jÞ respectively. From Eq. (1), it can be seen that the nonuniformity of the IRFPA is due to the deviation of the gain and the offset between detectors. The nonuniformity correction for the IRFPA is generally implemented in engineering by using the two-point correction algorithm. According to the radiation range of the scene which the IRFPA observes, we choose two irradiances u1 and u2 as the correction points, as shown in Fig. 2, and then calculate the average values of all N M pixel outputs Vij ðu1 Þ and Vij ðu2 Þ in the IRFPA respectively.
279
ϕ2
radiant intensity (w/m2)
Fig. 2. The linear model of response curve of detector in IRFPA.
V1 ¼
N X M X 1 Vij ðu1 Þ; N M i¼1 j¼1
N X M X 1 V2 ¼ Vij ðu2 Þ: N M i¼1 j¼1
ð2Þ
The line determined by ðVij ðu1 Þ; V 1 Þ and ðVij ðu2 Þ; V 2 Þ is used as the normalized line for the correction of the response curve of pixel ði; jÞ. Under real-time irradiance u, the output value Vij ðuÞ and the corrected value Vij0 ðu1 Þ have the relationship as follows Vij0 ðu1 Þ ¼
V2V1 Vij ðuÞ Vij ðu2 Þ Vij ðu1 Þ þV1
ðV 2 V 1 ÞVij ðu1 Þ ; Vij ðu2 Þ Vij ðu1 Þ
i ¼ 1; 2; . . . ; N ; j ¼ 1; 2; . . . ; M
ð3Þ
Let Gi;j ¼
V2V1 ; Vij ðu2 Þ Vij ðu1 Þ
ðV 2 V 1 ÞVij ðu1 Þ Oi;j ¼ V 1 Vij ðu2 Þ Vij ðu1 Þ
ð4Þ
Then Eq. (3) can be written as Vij0 ðu1 Þ ¼ Gij Vij ðuÞ þ Oij ;
ð5Þ
where Vi;j ð/Þ is the real-time output of the detector ði; jÞ under the incident irradiance u, Vi;j0 ð/Þ is the corrected value of Vi;j ð/Þ, and Gi;j and Oi;j are the correction coefficients for the gain and offset of the detector ði; jÞ respectively. Gi;j and Oi;j are precalculated by the DSP, and then stored in the FLASH. When the system is operating, the DSP
280
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
reads them out from the FLASH to correct realtime image data. 2.2.2. Algorithm for image enhancement and display How to effectively map 12 or 14 bit digitized image data to 8 bit TV video data is another important technology in IR image processing. The technology needs to reserve more information acquired from the field of view and accord with eye vision characteristics. The Histogram Equalization (HE) method is a simple and effective image enhancement and display algorithm[8,9]. HE expands the grey level that contains more pixels to attain more grey level. On the other hand, HE reduces the grey level that contains less pixels to obtain less grey levels. HE increases the information source entropy, so it increases the reserved amount of information acquired from the field of view. HE maps the grey scale of the input image to different grey scale. The map can be expressed as g ¼ T ðf Þ;
the IRFPA camera prototype. The IRFPA is InSb with 128 · 128 detectors, operating at a frame rate of 100 frames per second. Results are shown in Figs. 3 and 4 respectively. Fig. 3(a) is the raw image with nonuniformity. The thermal image of a hand is completely masked by the fixed pattern noise (nonuniformity). We cannot identify the image of the hand. The nonuniformity is 30%. Fig. 3(b) is the nonuniformity corrected image of Fig. 3(a). The hand emerges. The nonuniformity is reduced to 5%. NUC reduces the nonuniformity by a factor of 6. The imaging quality is greatly improved. Fig. 4(a) is the thermal image of the hand before enhancement processing. Fig. 4(b) is the image of image 4(a) after enhancement processing. It can be seen that the contrast and the information in the image are increased after enhancement and display processing.
ð6Þ
where f is the grey value of the input image with 0 6 f 6 L, where L is the total level of the grey of input image, and g is the grey value of the output image with 0 6 g 6 M, where M is the total level of the grey of the mapped image. HE makes the amount of pixel on every grey level approximately equal. The discrete expression for HE is gðf Þ ¼ ðM 1Þ
f f X nk ðM 1Þ X ¼ nk ; n n k¼0 k¼0
ð7Þ
Fig. 3. Effect of nonuniformity correction for the IRFPA: (a) raw thermal image of a hand with nonuniformity and (b) thermal image of a hand after nonuniformity correction.
where nk is the pixel amount on grey level k and n is the pixel amount of the input image. The implementation of bit mapping is achieved by using a Look-up Table (LUT). Statistics of the histogram and the computation for the relationship of bit mapping are accomplished in the FPGA.
3. Experimental results The performance and capabilities of the IRFPA signal processing system are validated by procedures that connect the signal processing system to
Fig. 4. Effect of enhancement to infrared imaging of a hand. The thermal image of hand (a) before enhancement and (b) after enhancement.
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
The experimental results show that the system presented has the function of real-time processing IRFPA imaging signal, which can meet the requirements for processing more complicated IRFPA imaging signals. The system also has the advantages of high speed, high precision, powerful processing capability, and large flexibility.
4. Conclusions The application of IR imaging technology will be enlarged continuously, according to which, the requirement for the ability of the imaging signal processing system to process information will be higher and higher. The proposed real-time signal processing system based on DSP&FPGA, which can be expanded flexibly and which can easily implement different level processing algorithms, can fulfill these complex tasks. So it will be an important developing trend of the IR imaging signal processing system. Moreover, nonuniformity correction and image enhancement and display technology are also the key and interesting tasks of the IRPFA imaging system.
281
References [1] D.A. Scribner, M.R. Kruer, C.J. Gridley, Measurement, characterization, and modeling of noise in staring infrared focal plan array, SPIE 782 (1987) 147–160. [2] Z. Cao, N. Sang, Digital implemention of nonuniformity correction for IRFPAs, SPIE 3698 (1999) 807–814. [3] S.R. Horman, K.C. hepfer, M. Zurasky, Uniformity compensation for high quantum efficiency focal arrays, SPIE 2744 (1996) 154–164. [4] A.F. Milton, F.R. Barone, M.R. Kruer, Influence of nonuniformity on infrared focal plane array, Opt. Eng. 24 (5) (1985) 855–862. [5] D.A. Scribner, M.R. Kruer, C.J. Gridley, Physical limitation to nonuniformity correction in IR focal plane arrays, SPIE 865 (1987) 185–202. [6] M. Schulz, L. Caldwell, Nonuniformity correction and correctability of infrared plane arrays, SPIE 2470 (1995) 200–209. [7] L.P. David, L.D. Eustace, Linear theory of nonuniformity correction in infrared staring sensors, Opt. Eng. 32 (8) (1993) 1854–1859. [8] J. Silverman, Signal processing algorithms for display and enhancement of IR images, SPIE 2020 (1993) 440– 450. [9] J. Woodruff Christopher, Efficient display of high-dynamic-range imagery, SPIE 2552 (1995) 56–64. [10] G. Jiang, S. Liu, Adaptive nonuniformity correction of IRFPA, J. Infrared Millim. Waves 20 (2) (2001) 93–96 (Chinese).
A new real-time processing system for the IRFPA imaging signal based on DSP&FPGA Hui-Xin Zhou *, Rui Lai, Shang-Qiang Liu, Bing-Jian Wang, Qing Li Lab. of Infrared Technology, School of Technical Physics, P.O. Box 204, Xidian University, Xi’an 710071, PR China Received 10 December 2003 Available online 6 July 2004
Abstract The principle, configuration and special features of a new high-precision real-time signal processing system for Infrared Focal Plane Arrays is presented in this note. A structure based on DSP&FPGA is adopted in the system. The FPGA implements the system timing control and the low level algorithms. The DSP performs the high-level algorithms, such as nonuniformity correction for the IRFPA. Nonuniformity correction and image enhancement and display algorithm, the two crucial algorithms for IR imaging signal processing, are also discussed in this note. The experiments on real IR imaging sequences demonstrate that the system is suitable for the real-time high-speed IR imaging system with high quality and high precision. 2004 Elsevier B.V. All rights reserved. PACS: 07.57.)c; 07.57.kp; 42.30.va Keywords: Infrared Focal Plan Arrays; Infrared imaging; Nonuniformity correction; Image processing; DSP; FPGA
1. Introduction With the development of Infrared focal plane array (IRFPA) technology, the advantages of high density, excellent performance, high reliability and miniaturization are available in Infrared (IR) imaging systems [1]. At present, acquisition of high quality images has become the key problem of IR imaging systems. Such systems generally need to process mass data in real-time [2]. The processing
*
Corresponding author. E-mail addresses: [email protected] Zhou), [email protected] (R. Lai).
(H.-X.
includes various algorithms, such as nonuniformity correction, image segmentation, local characteristics extraction, image de-noising, image enhancement, etc. hence then must be one well integrated high-speed information processing system. In this paper a new IRFPA imaging system based on the DSP&FPGA is presented to fulfill such requirements. Our note also focuses on two algorithms that are used in the IR image processing stage. One is the nonuniformity correction and the other is image enhancement and display method. Nonuniformity correction is a key problem that must be first solved in IRFPA systems. The detector-to-detector responsivity (gain) and dark current (offset)
1350-4495/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.infrared.2004.02.003
278
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
variations [3–7], which cause the nonuniformity, may completely mask the useful thermal signatures in an IR image with a fixed pattern ‘‘noise’’. In order to get better resolution, a 12 or 14 bit ADC is commonly adopted in the signal processing system [2]. How to effectively map 12 or 14 bit data to 8 bit TV video data is another important technology in IR image processing systems [8,9].
2. New real-time processing system for IRFPA imaging signal 2.1. Hardware configuration of the system The schematic diagram of the real-time signal processing system for IRFPA imaging based on DSP&FPGA is shown in Fig. 1. The main function unit of the system consists of an IRFPA driving circuit, an analog imaging signal processor, an ADC, a high-speed signal processor, a DAC and a video synthesizer, etc. The IRFPA driving circuit unit provides the pulse-driving signal and the offset voltage, which are crucial for the IRFPA operating state. It makes the IRFPA work at the optimum operating point. In the traditional method, the driving circuit would be complex and nonflexible. In our system, the driving-pulse signal is achieved by an FPGA. The method has the advantages of high reliability and flexibility. The pre-processing circuit unit amplifies the imaging analog signal with low noise to the level that the ADC requires. The ADC transforms the analog image signal to a digital one. In order to be
applied to image data processing with high speed and precision, a 12 bit ADC whose sampling frequency is up to 20 MHz is selected, so that a high resolution of digitized image data is obtained. The synchronization and timing control unit harmonizes the other units in the system, including the output circuit unit of the IRFPA, the ADC sample unit, the data store, the DSP, the DAC, the video synthesizer, etc. The DSP&FPGA unit is the kernel of the system. In the system of real-time IR signal processing, the low-level processing algorithms that only need simple computation deal with a large amount of data. But they do not need high-processing speed. The FPGA is adapted to doing these, and flexibility is also achieved. The high-level algorithms process smaller amounts of data, but are more complex. The DSP (ADSP21060) has the characteristics of high-speed operation, flexible addressing method and powerful communication capability. Thus the DSP is perfect for implementing the high-level algorithm. The memory must be expanded for storing a large amount of IRFPA imaging data. In this note, SRAM is used as expanded data memory, which has the advantages of fast access and lager capacity in a single chip. As an embedded DSP system, the off-chip program memory is also needed. FLASH memory is the preferred choice, which can do online erasing and programming. It also enhances the flexibility of the system. The image data processed by the DSP&FPGA module are sent to the DAC through a data buffer, where the digital image data are transformed back to an analog image signal. Under the Synchroni-
RESET
IRFPA
Preprocessing
ADC
FPGA
DSP
SRAM
Driving Circuit
FLASH
Data Buffer
DAC
Video Synthesizer
Synchronization and Timing Control Module
Fig. 1. Schematic diagram of the signal processing system for IRFPA imaging.
Monitor
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
zation signal, the analog image signal is synthesized to standard video signal in the Video synthesizer unit. Finally the image is displayed in real-time on the monitor.
1 2
gray level
2.2. Software for the algorithm
3
ϕ1
Using the technology of DSP&FPGA, with a real-time requirement, we can adopt different processing algorithms to different IR imaging signals. In the limited space of the note, only two crucial algorithms for IR imaging signal processing are discussed. 2.2.1. Nonuniformity correction algorithm The so-called nonuniformity of IRFPA is the variation of response output between the detectors in the IRFPA under uniform background illumination. There are several factors causing nonuniformity. The main sources of nonuniformity are: (1) responsivity nonuniformity, including spectral response nonuniformity; (2) nonuniformity of the readout circuit and the coupling between the detector and the readout circuit; and (3) nonuniformity of dark current [4]. Without nonuniformity compensation (NUC), the images from the IRFPA are distorted and are not suitable for field use. The individual detector outputs are linear and stable in time, so they can be expressed as [3–7,10] Vi;j ðuÞ ¼ Ri;j ðuÞu þ vi;j ðuÞ;
ð1Þ
where ði; jÞ are the coordinates of a detector in the array, u represents the irradiance that is incident on the detector ði; jÞ, Vi;j ðuÞ is the output of the detector ði; jÞ, and Ri;j ðuÞ and vi;j ðuÞ are the gain and the offset of the detector ði; jÞ respectively. From Eq. (1), it can be seen that the nonuniformity of the IRFPA is due to the deviation of the gain and the offset between detectors. The nonuniformity correction for the IRFPA is generally implemented in engineering by using the two-point correction algorithm. According to the radiation range of the scene which the IRFPA observes, we choose two irradiances u1 and u2 as the correction points, as shown in Fig. 2, and then calculate the average values of all N M pixel outputs Vij ðu1 Þ and Vij ðu2 Þ in the IRFPA respectively.
279
ϕ2
radiant intensity (w/m2)
Fig. 2. The linear model of response curve of detector in IRFPA.
V1 ¼
N X M X 1 Vij ðu1 Þ; N M i¼1 j¼1
N X M X 1 V2 ¼ Vij ðu2 Þ: N M i¼1 j¼1
ð2Þ
The line determined by ðVij ðu1 Þ; V 1 Þ and ðVij ðu2 Þ; V 2 Þ is used as the normalized line for the correction of the response curve of pixel ði; jÞ. Under real-time irradiance u, the output value Vij ðuÞ and the corrected value Vij0 ðu1 Þ have the relationship as follows Vij0 ðu1 Þ ¼
V2V1 Vij ðuÞ Vij ðu2 Þ Vij ðu1 Þ þV1
ðV 2 V 1 ÞVij ðu1 Þ ; Vij ðu2 Þ Vij ðu1 Þ
i ¼ 1; 2; . . . ; N ; j ¼ 1; 2; . . . ; M
ð3Þ
Let Gi;j ¼
V2V1 ; Vij ðu2 Þ Vij ðu1 Þ
ðV 2 V 1 ÞVij ðu1 Þ Oi;j ¼ V 1 Vij ðu2 Þ Vij ðu1 Þ
ð4Þ
Then Eq. (3) can be written as Vij0 ðu1 Þ ¼ Gij Vij ðuÞ þ Oij ;
ð5Þ
where Vi;j ð/Þ is the real-time output of the detector ði; jÞ under the incident irradiance u, Vi;j0 ð/Þ is the corrected value of Vi;j ð/Þ, and Gi;j and Oi;j are the correction coefficients for the gain and offset of the detector ði; jÞ respectively. Gi;j and Oi;j are precalculated by the DSP, and then stored in the FLASH. When the system is operating, the DSP
280
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
reads them out from the FLASH to correct realtime image data. 2.2.2. Algorithm for image enhancement and display How to effectively map 12 or 14 bit digitized image data to 8 bit TV video data is another important technology in IR image processing. The technology needs to reserve more information acquired from the field of view and accord with eye vision characteristics. The Histogram Equalization (HE) method is a simple and effective image enhancement and display algorithm[8,9]. HE expands the grey level that contains more pixels to attain more grey level. On the other hand, HE reduces the grey level that contains less pixels to obtain less grey levels. HE increases the information source entropy, so it increases the reserved amount of information acquired from the field of view. HE maps the grey scale of the input image to different grey scale. The map can be expressed as g ¼ T ðf Þ;
the IRFPA camera prototype. The IRFPA is InSb with 128 · 128 detectors, operating at a frame rate of 100 frames per second. Results are shown in Figs. 3 and 4 respectively. Fig. 3(a) is the raw image with nonuniformity. The thermal image of a hand is completely masked by the fixed pattern noise (nonuniformity). We cannot identify the image of the hand. The nonuniformity is 30%. Fig. 3(b) is the nonuniformity corrected image of Fig. 3(a). The hand emerges. The nonuniformity is reduced to 5%. NUC reduces the nonuniformity by a factor of 6. The imaging quality is greatly improved. Fig. 4(a) is the thermal image of the hand before enhancement processing. Fig. 4(b) is the image of image 4(a) after enhancement processing. It can be seen that the contrast and the information in the image are increased after enhancement and display processing.
ð6Þ
where f is the grey value of the input image with 0 6 f 6 L, where L is the total level of the grey of input image, and g is the grey value of the output image with 0 6 g 6 M, where M is the total level of the grey of the mapped image. HE makes the amount of pixel on every grey level approximately equal. The discrete expression for HE is gðf Þ ¼ ðM 1Þ
f f X nk ðM 1Þ X ¼ nk ; n n k¼0 k¼0
ð7Þ
Fig. 3. Effect of nonuniformity correction for the IRFPA: (a) raw thermal image of a hand with nonuniformity and (b) thermal image of a hand after nonuniformity correction.
where nk is the pixel amount on grey level k and n is the pixel amount of the input image. The implementation of bit mapping is achieved by using a Look-up Table (LUT). Statistics of the histogram and the computation for the relationship of bit mapping are accomplished in the FPGA.
3. Experimental results The performance and capabilities of the IRFPA signal processing system are validated by procedures that connect the signal processing system to
Fig. 4. Effect of enhancement to infrared imaging of a hand. The thermal image of hand (a) before enhancement and (b) after enhancement.
H.-X. Zhou et al. / Infrared Physics & Technology 46 (2005) 277–281
The experimental results show that the system presented has the function of real-time processing IRFPA imaging signal, which can meet the requirements for processing more complicated IRFPA imaging signals. The system also has the advantages of high speed, high precision, powerful processing capability, and large flexibility.
4. Conclusions The application of IR imaging technology will be enlarged continuously, according to which, the requirement for the ability of the imaging signal processing system to process information will be higher and higher. The proposed real-time signal processing system based on DSP&FPGA, which can be expanded flexibly and which can easily implement different level processing algorithms, can fulfill these complex tasks. So it will be an important developing trend of the IR imaging signal processing system. Moreover, nonuniformity correction and image enhancement and display technology are also the key and interesting tasks of the IRPFA imaging system.
281
References [1] D.A. Scribner, M.R. Kruer, C.J. Gridley, Measurement, characterization, and modeling of noise in staring infrared focal plan array, SPIE 782 (1987) 147–160. [2] Z. Cao, N. Sang, Digital implemention of nonuniformity correction for IRFPAs, SPIE 3698 (1999) 807–814. [3] S.R. Horman, K.C. hepfer, M. Zurasky, Uniformity compensation for high quantum efficiency focal arrays, SPIE 2744 (1996) 154–164. [4] A.F. Milton, F.R. Barone, M.R. Kruer, Influence of nonuniformity on infrared focal plane array, Opt. Eng. 24 (5) (1985) 855–862. [5] D.A. Scribner, M.R. Kruer, C.J. Gridley, Physical limitation to nonuniformity correction in IR focal plane arrays, SPIE 865 (1987) 185–202. [6] M. Schulz, L. Caldwell, Nonuniformity correction and correctability of infrared plane arrays, SPIE 2470 (1995) 200–209. [7] L.P. David, L.D. Eustace, Linear theory of nonuniformity correction in infrared staring sensors, Opt. Eng. 32 (8) (1993) 1854–1859. [8] J. Silverman, Signal processing algorithms for display and enhancement of IR images, SPIE 2020 (1993) 440– 450. [9] J. Woodruff Christopher, Efficient display of high-dynamic-range imagery, SPIE 2552 (1995) 56–64. [10] G. Jiang, S. Liu, Adaptive nonuniformity correction of IRFPA, J. Infrared Millim. Waves 20 (2) (2001) 93–96 (Chinese).