System and method for making colored low-light image

Abstract

The invention relates to a system and a method for making a colored low-light image. The system for making the color low-light-level image comprises a high-sensitivity CMOS imaging module which is used for enhancing the photon collection capability and reducing the noise level to complete imaging; the image enhancement processing module adopts a histogram equalization algorithm and comprises the steps of calculating a gray histogram of the original image, calculating a cumulative distribution function according to the gray histogram and performing gray mapping to obtain an enhanced gray value; the image noise reduction processing module is used for carrying out noise reduction on the image after enhancement processing, and filtering is carried out by adopting a night image noise reduction technology, a bilateral filtering method including spatial domain noise reduction and a band-pass filter including frequency domain noise reduction; and the colorful low-light-level image generation module obtains a colorful low-light-level night vision image by adopting an interpolation algorithm and makes a colorful low-light-level image. According to the system and the method for manufacturing the colored low-light-level image, special equipment is not depended on, the imaging quality of the colored low-light-level image can be improved, and the colored low-light-level image can be manufactured.

Description

Technical Field

[0001] The present invention relates to the technical field of image data analysis of optoelectronic aiming sights, and more particularly, to a system and method for producing color low-light images. Background Art

[0002] For image restoration in low-light scenarios, there are currently two main research areas: the hardware area and the algorithm area. In the hardware area, researchers are committed to designing circuit devices to obtain richer photosensitive information and overcome device noise problems; while in the algorithm area, focus is on processing the received information to restore scene information. The research in these two areas complements each other and is indispensable. However, currently, most images in low-light scenarios need to be taken through devices such as image intensifiers and thermal imagers. These devices are not only expensive but also unable to truly restore the color information of the original scene. Therefore, while improving the performance of dedicated imaging devices and reducing costs, it is of great research significance to study low-light scene image enhancement algorithms that do not rely on dedicated devices.

[0003] In daily life, we often encounter images with overall or partial low brightness, such as at night, on rainy days, insufficient lighting, and under-exposed cameras. In these cases, the images usually show insufficient overall brightness and contrast, and may contain noise. Especially for low-light images taken in extremely low-light environments, the noise contained is obvious, resulting in poor visual effects and affecting information transmission.

[0004] Although there are already a variety of classic low-light enhancement algorithms, they generally have problems with low imaging quality such as over-enhancement, color distortion, and detail loss. In addition, these algorithms usually only enhance the image brightness, while ignoring the amplified noise components during the enhancement process, resulting in poor visual effects of the enhanced images and affecting the restoration of the original scene detail information of the images. At the same time, during the image acquisition process, due to uneven lighting in the shooting environment, there may be local areas with uneven exposure or under-exposure, and the detail information in these areas is often difficult to distinguish with the naked eye, resulting in insufficient information. In order to make the image lighting more uniform and highlight more detail information in dark areas, it is necessary to correct such images with uneven lighting. The processing method focuses more on image lighting balance compared to low-light image enhancement, that is, performing brightness balance processing on different lighting brightness regions.

[0005] Therefore, there is an urgent need to develop a method that does not rely on dedicated devices and improves the imaging quality of color low-light images. Summary of the Invention

[0006] The technical problem to be solved by the present invention is the problem of low imaging quality of low-light images in the prior art, and how to provide a method for producing color low-light images that does not rely on dedicated devices.

[0007] To solve the above technical problems, according to one aspect of the present invention, a system for making a color low-light image is provided, which includes: a high-sensitivity CMOS (Complementary Metal Oxide Semiconductor) imaging module. The high-sensitivity CMOS imaging module is an image sensor, which is used for enhanced photon collection ability and reduced noise level. The received light signal is converted into a charge signal through the built-in micro-photodiode array, and is amplified, filtered and digitally processed by the internal circuit to be converted into digital image data to complete imaging; an image enhancement processing module, which is used for enhancing the imaged image. The histogram equalization algorithm is adopted, including calculating the gray histogram of the original image, calculating the cumulative distribution function according to the gray histogram, and performing gray mapping to obtain the enhanced gray value; an image noise reduction processing module, which is used for reducing the noise of the enhanced image. The night image noise reduction technology is adopted, and the bilateral filtering method including spatial domain noise reduction and the band-pass filter for frequency domain noise reduction are used for filtering. Among them, the bilateral filtering method for spatial domain noise reduction combines the spatial domain and gray domain information to determine the filtering weight, and can better retain the edges and details of the image while removing noise; the design of the band-pass filter adopts the method of Butterworth filter or Chebyshev filter. First, the night image is converted from the spatial domain to the frequency domain, the noise frequency range and the frequency range where the image detail information is located are determined, and after designing and applying the band-pass filter, the inverse transformation is performed to obtain the noise-reduced image in the spatial domain; a color low-light image generation module. The color low-light image generation module adopts an interpolation algorithm to obtain a color low-light night vision image. The low-light image is converted from the RGB space to the YUV space, and the missing color components in the YUV image are estimated by using the bilinear interpolation algorithm. The value of each pixel point is calculated by linearly weighted averaging the values of the four known pixel points around the pixel point. Finally, the image in the processed YUV color space is converted back to the RGB color space to obtain a color low-light image, completing the production of the color low-light image.

[0008] According to an embodiment of the present invention, the sensor of the high-sensitivity CMOS image module can maintain high image clarity and signal-to-noise ratio under extremely low light conditions, and is applicable to specific application scenarios such as night, dim environment or light-limited conditions. Among them, the high-sensitivity CMOS image module can perform preprocessing on the image, including automatic exposure control AEC (Automatic Exposure Control) and automatic gain control AGC (Automatic Gain Control), so as to dynamically adjust the exposure time and gain setting according to the change of ambient light, ensure the best imaging effect under different light conditions, and at the same time, through the built-in or external image processing module, preliminary denoising and white balance processing can be performed.

[0009] According to an embodiment of the present invention, the image enhancement processing module may adopt a night vision image enhancement algorithm to improve the quality of images acquired under low light conditions for better observation and analysis of targets. The night vision image enhancement algorithm uses the histogram equalization algorithm for image enhancement processing, including the following steps: First, calculate the grayscale histogram of the original image. Traverse each pixel of the original image and count the number of occurrences of each grayscale level to obtain the grayscale histogram of the original image. For an image of size M×N, whose grayscale range is from 0 to L-1, the grayscale histogram h(k) represents the number of occurrences of grayscale level k, where k = 0, 1, … L-1. Second, calculate the cumulative distribution function c(k) according to the grayscale histogram. The formula is: where k = 0, 1, … L-1, and c(k) represents the total number of pixels with grayscale levels less than or equal to k. Finally, perform grayscale mapping, normalize the cumulative distribution function to obtain the mapped grayscale value. The normalization formula is: where M×N is the total number of pixels in the image and L is the total number of grayscale levels. For each pixel in the original image, according to its grayscale value k, obtain the enhanced grayscale value through the mapping relationship s(k).

[0010] According to an embodiment of the present invention, the image noise reduction processing module may adopt night image noise reduction technology to reduce noise in the enhanced image, including spatial domain noise reduction, frequency domain noise reduction, and 3D noise reduction methods to reduce noise in the enhanced image. Among them, spatial domain noise reduction is a filtering method that operates on pixels in the image space. By performing weighted summation on each pixel in the image and its neighboring pixels, the value of the pixel is changed. The bilateral filtering method is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and grayscale domain information. By considering the spatial distance and grayscale difference between pixel points to determine the filtering weight, it can better retain the edges and details of the image while removing noise. The calculation formula for bilateral filtering is: where, I out (p) is the grayscale value of the output image at pixel point p, W p is the normalization factor, S is the neighborhood, is the spatial domain Gaussian function, is the Gaussian function in the gray scale domain, ||p - q|| is the spatial distance between pixel points, and ||I(p) - I(q)|| is the gray scale difference between pixel points; among them, after performing spatial domain noise reduction, frequency domain noise reduction is then adopted to filter out salt-and-pepper noise, Gaussian noise, and Poisson noise in the night vision image. In frequency domain filtering, a band-pass filter is used for filtering. First, the night image is converted from the spatial domain to the frequency domain by using the method of Fourier transform or wavelet transform; then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range where the detailed information of the image is located; according to the analysis results, appropriate band-pass filter parameters are selected to design the band-pass filter; among them, the band-pass filter adopts Butterworth filter and Chebyshev filter, which have a smooth transition band and good filtering performance; the designed band-pass filter is applied to the image in the frequency domain to perform filtering processing on the image; the filtering processing includes direct filtering method or indirect filtering method. Among them, the direct filtering method multiplies the frequency response of the filter by the frequency spectrum of the image to obtain the filtered frequency spectrum, and then obtains the filtered spatial domain image through inverse transformation; the indirect filtering method first performs inverse transformation on the filter to obtain the filter in the spatial domain, then convolves the filter with the image in the spatial domain to obtain the filtered image, and finally performs inverse transformation on the filtered frequency domain image to obtain the noise-reduced image in the spatial domain.

[0011] According to an embodiment of the present invention, the color low-light image generation module can use an interpolation algorithm to obtain a color low-light night vision image. First, the low-light image is converted from the RGB space to the YUV space, and the conversion formulas are respectively: Y = 0.299R + 0.587G + 0.114B; U = -0.147R - 0.289G + 0.436B; 0.615R - 0.515G - 0.100B, where R, G, and B respectively represent the red, green, and blue components in the RGB color space, and Y, U, and V respectively represent the luminance component and chrominance components in the YUV color space; then, the value of the unknown pixel is estimated by weighted averaging of adjacent pixels. During the process of converting a black-and-white image to a color image, bilinear interpolation algorithm is used to estimate the missing color components; among them, for each pixel point in the YUV image, the value of the pixel point is calculated by linearly weighting and averaging the values of its surrounding four known pixel points; finally, the image in the processed YUV color space is converted back to the RGB color space to obtain the color low-light image.

[0012] According to another aspect of the present invention, there is provided a method for producing a color low-light image. The method for producing a color low-light image is implemented based on a system for producing a color low-light image. The system for producing a color low-light image includes: a high-sensitivity CMOS imaging module, an image enhancement processing module, an image noise reduction processing module, and a color low-light image generation module. The method for producing a color low-light image includes the following steps: S1. High-sensitivity CMOS imaging is achieved through the high-sensitivity CMOS imaging module to enhance the photon collection ability and reduce the noise level. The received light signal is converted into a charge signal by the built-in micro-photodiode array, and after being amplified, filtered, and digitized by the internal circuit, it is converted into digital image data to complete the imaging; S2. Image enhancement processing is performed on the imaged image. The histogram equalization algorithm is adopted, including calculating the gray histogram of the original image, calculating the cumulative distribution function according to the gray histogram, and performing gray mapping to obtain the enhanced gray value; S3. Image noise reduction processing is performed on the image after the enhancement processing. The night image noise reduction technology is adopted, and the bilateral filtering method including spatial domain noise reduction and the band-pass filter for frequency domain noise reduction are used for filtering. Among them, the bilateral filtering method for spatial domain noise reduction combines the spatial domain and gray domain information to determine the filtering weight, and can better retain the edges and details of the image while removing the noise; the design of the band-pass filter adopts the method of Butterworth filter or Chebyshev filter. First, the night image is converted from the spatial domain to the frequency domain, the noise frequency range and the frequency range where the image detail information is located are determined, and after designing and applying the band-pass filter, the inverse transformation is performed to obtain the noise-reduced image in the spatial domain; S4. Color low-light image generation. The interpolation algorithm is used to obtain the color low-light night vision image. The low-light image is converted from the RGB space to the YUV space, and the missing color components in the YUV image are estimated by using the bilinear interpolation algorithm. The value of each pixel point is calculated by linearly weighting and averaging the values of the four known pixel points around the pixel point. Finally, the image in the processed YUV color space is converted back to the RGB color space to obtain the color low-light image, and the production of the color low-light image is completed.

[0013] According to an embodiment of the present invention, in step S1, the high-sensitivity CMOS image module can perform preprocessing on the image, including automatic exposure control (AEC) and automatic gain control (AGC), to dynamically adjust the exposure time and gain setting according to the change of the ambient light, ensuring the best imaging effect under different lighting conditions. At the same time, through the built-in or external image processing module, preliminary noise reduction and white balance processing can be performed.

[0014] According to an embodiment of the present invention, in step S1, the image enhancement process may adopt a night vision image enhancement algorithm to improve the quality of images acquired under low light conditions for better observation and analysis of targets. The night vision image enhancement algorithm uses the histogram equalization algorithm for image enhancement processing, including the following steps: First, calculate the grayscale histogram of the original image. Traverse each pixel of the original image and count the number of occurrences of each grayscale level to obtain the grayscale histogram of the original image. For an image of size M×N, the range of its grayscale levels is from 0 to L-1, then the grayscale histogram h(k) represents the number of occurrences of grayscale level k, where k = 0, 1, … L-1; Second, calculate the cumulative distribution function c(k) according to the grayscale histogram. The formula is: where k = 0, 1, … L-1, and c(k) represents the total number of pixels with grayscale levels less than or equal to k; Finally, perform grayscale mapping, normalize the cumulative distribution function to obtain the mapped grayscale value. The normalization formula is: where M×N is the total number of pixels in the image and L is the total number of grayscale levels; For each pixel in the original image, according to its grayscale value k, obtain the enhanced grayscale value through the mapping relationship s(k).

[0015] According to an embodiment of the present invention, in step S3, the image noise reduction process may adopt a night image noise reduction technology to perform noise reduction on the enhanced image, including spatial domain noise reduction, frequency domain noise reduction, and 3D noise reduction methods to perform noise reduction on the enhanced image. Among them, spatial domain noise reduction is a filtering method that operates on pixels in the image space. By performing weighted summation on each pixel in the image and its neighboring pixels, the value of the pixel is changed; The bilateral filtering method is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and grayscale domain information. By considering the spatial distance and grayscale difference between pixel points to determine the filtering weight, it can better retain the edges and details of the image while removing noise. The calculation formula for bilateral filtering is: where, I out (p) is the grayscale value of the output image at pixel point p, W p is the normalization factor, S is the neighborhood, is the spatial domain Gaussian function, is the Gaussian function in the grayscale domain, ||p - q|| is the spatial distance between pixel points, and ||I(p) - I(q)|| is the grayscale difference between pixel points; among them, after performing spatial domain noise reduction and then frequency domain noise reduction, salt-and-pepper noise, Gaussian noise, and Poisson noise in the night vision image are filtered out. In frequency domain filtering, a band-pass filter is used for filtering. First, the night image is converted from the spatial domain to the frequency domain by using the method of Fourier transform or wavelet transform; then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range where the detailed information of the image is located; according to the analysis results, appropriate band-pass filter parameters are selected to design the band-pass filter; among them, the band-pass filter uses Butterworth filter and Chebyshev filter, which have a smooth transition band and good filtering performance; the designed band-pass filter is applied to the image in the frequency domain to perform filtering on the image; the filtering process includes direct filtering method or indirect filtering method. Among them, the direct filtering method multiplies the frequency response of the filter by the frequency spectrum of the image to obtain the filtered frequency spectrum, and then obtains the filtered spatial domain image through inverse transformation; the indirect filtering method first performs inverse transformation on the filter to obtain the filter in the spatial domain, then performs convolution operation on the filter and the image in the spatial domain to obtain the filtered image, and finally, performs inverse transformation on the filtered frequency domain image to obtain the noise-reduced image in the spatial domain.

[0016] According to an embodiment of the present invention, in step S4, the generation of the color low-light image can use an interpolation algorithm to obtain the color low-light night vision image. First, the low-light image is converted from the RGB space to the YUV space, and the conversion formulas are respectively: Y = 0.299R + 0.587G + 0.114B; U = -0.147R - 0.289G + 0.436B; 0.615R - 0.515G - 0.100B, where R, G, and B respectively represent the red, green, and blue components in the RGB color space, and Y, U, and V respectively represent the luminance component and chrominance component in the YUV color space; then, the value of the unknown pixel is estimated by weighted averaging of adjacent pixels. In the process of converting a black-and-white image to a color image, the bilinear interpolation algorithm is used to estimate the missing color components; among them, for each pixel point in the YUV image, the value of the pixel point is calculated by linearly weighting and averaging the values of its surrounding four known pixel points; finally, the image in the processed YUV color space is converted back to the RGB color space to obtain the color low-light image.

[0017] Compared with the prior art, the technical solutions provided by the embodiments of the present invention can at least achieve the following beneficial effects:

[0018] The system and method for making color low-light images according to the present invention do not rely on dedicated equipment and can improve the imaging quality of color low-light images, thereby realizing the production of color low-light images. The present invention uses a high-sensitivity CMOS as the color night vision module and adopts a night vision ISP algorithm to achieve color imaging under an extremely low illuminance of 0.001 lux. Different from traditional infrared thermal imaging and image intensifier tubes, the color night vision module utilizes the visible light band of the human eye, and the imaging effect is closer to what the human eye sees.

[0019] The system and method for making color low-light images according to the present invention use a high-sensitivity CMOS sensor, which has significantly enhanced photon collection ability and lower noise level compared with traditional sensors. This enables it to maintain high image clarity and signal-to-noise ratio even under extremely low light conditions. This characteristic is particularly important for night, dim environments or specific application scenarios with limited light, which can significantly improve the imaging quality and capture more details that are difficult to detect by the naked eye.

[0020] In the spatial domain noise reduction of the system and method for making color low-light images according to the present invention, the bilateral filtering method is adopted to determine the filtering weights by combining spatial domain and gray scale domain information, which can better retain the edge and detail information of the image while removing noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present invention and do not limit the present invention.

[0022] Figure 1 It is a flowchart showing the operation of the system and method for making color low-light images according to the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0024] Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings as understood by those of ordinary skill in the art to which this invention pertains. The terms "first", "second" and similar terms used in the description and claims of this patent application for invention do not denote any order, quantity or importance, but are only used to distinguish different components. Similarly, terms such as "a" or "an" do not denote a limitation of quantity, but mean that there is at least one.

[0025] Figure 1 is a flowchart showing the operation of a system and method for producing a color low-light image according to an embodiment of the present invention.

[0026] As Figure 1 shown, the system for producing a color low-light image includes: a high-sensitivity CMOS imaging module, an image enhancement processing module, an image noise reduction processing module, and a color low-light image generation module.

[0027] The high-sensitivity CMOS imaging module is an image sensor, which is used to enhance the photon collection ability and reduce the noise level. It converts the received light signal into a charge signal through an internal micro photosensitive diode array, and after amplification, filtering and digital processing by an internal circuit, it is converted into digital image data to complete imaging.

[0028] The image enhancement processing module is used to enhance the imaged image. It adopts the histogram equalization algorithm, including calculating the gray histogram of the original image, calculating the cumulative distribution function according to the gray histogram, and performing gray mapping to obtain the enhanced gray value.

[0029] The image noise reduction processing module is used to reduce the noise of the image after enhancement processing. It adopts the night image noise reduction technology, and uses the bilateral filtering method for spatial domain noise reduction and the band-pass filter for frequency domain noise reduction for filtering. Among them, the bilateral filtering method for spatial domain noise reduction combines the spatial domain and gray domain information to determine the filtering weight, and can better retain the edges and details of the image while removing noise; the design of the band-pass filter adopts the method of Butterworth filter or Chebyshev filter. First, the night image is converted from the spatial domain to the frequency domain, the noise frequency range and the frequency range where the image details are located are determined, and after designing and applying the band-pass filter, the inverse transformation is performed to obtain the noise-reduced image in the spatial domain.

[0030] The color low-light image generation module uses the interpolation algorithm to obtain the color low-light night vision image. It converts the low-light image from the RGB space to the YUV space, estimates the missing color components in the YUV image by using the bilinear interpolation algorithm, calculates the value of each pixel point by linearly weighting and averaging the values of the four known pixel points around the pixel point, and finally converts the processed image in the YUV color space back to the RGB color space to obtain the color low-light image, completing the production of the color low-light image.

[0031] The system and method for making color low-light images according to the present invention can achieve the production of color low-light images without relying on special equipment. The present invention uses a high-sensitivity CMOS as the color night vision core and adopts a night vision ISP algorithm to achieve color imaging under an extremely low illuminance of 0.001 lux. Different from traditional infrared thermal imaging and image intensifier tubes, the color night vision core utilizes the visible light band of the human eye, and the imaging effect is closer to what the human eye sees.

[0032] According to one or some embodiments of the present invention, the sensor of the high-sensitivity CMOS image module can maintain high image clarity and signal-to-noise ratio under extremely low light conditions, and is suitable for specific application scenarios at night, in dim environments or with limited light. Among them, the high-sensitivity CMOS image module preprocesses the image, including automatic exposure control (AEC) and automatic gain control (AGC), to dynamically adjust the exposure time and gain settings according to the change of ambient light, ensuring the best imaging effect under different light conditions. At the same time, through the built-in or external image processing module, preliminary denoising and white balance processing can be performed.

[0033] According to one or some embodiments of the present invention, the image enhancement processing module adopts a night vision image enhancement algorithm to improve the image quality obtained under low light conditions for better observation and analysis of the target. The night vision image enhancement algorithm uses the histogram equalization algorithm for image enhancement processing, including the following steps:

[0034] First, calculate the gray histogram of the original image. Traverse each pixel of the original image and count the number of occurrences of each gray level to obtain the gray histogram of the original image. For an image of size M×N, the gray level range is from 0 to L - 1, then the gray histogram h(k) represents the number of occurrences of gray level k, where k = 0, 1,..., L - 1.

[0035] Secondly, calculate the cumulative distribution function c(k) according to the gray histogram. The formula is: where k = 0, 1,..., L - 1, and c(k) represents the total number of pixels with gray levels less than or equal to k.

[0036] Finally, perform gray mapping, normalize the cumulative distribution function to obtain the mapped gray value. The normalization formula is: where M×N is the total number of pixels in the image, and L is the total number of gray levels. For each pixel in the original image, according to its gray value k, the enhanced gray value is obtained through the mapping relationship s(k).

[0037] According to one or some embodiments of the present invention, the image noise reduction processing module uses night image noise reduction technology to reduce the noise of the enhanced image, including spatial domain noise reduction, frequency domain noise reduction, and 3D noise reduction methods to reduce the noise of the enhanced image.

[0038] Among them, spatial domain noise reduction is a filtering method that operates on pixels in the image space. By performing weighted summation on each pixel in the image and its neighboring pixels, the value of the pixel is changed; the bilateral filtering method is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and gray domain information. By considering the spatial distance and gray difference between pixel points to determine the filtering weight, it can better retain the edges and details of the image while removing noise; the calculation formula of bilateral filtering is: Among them, I out (p) is the gray value of the output image at pixel point p, W p is the normalization factor, S is the neighborhood, is the spatial domain Gaussian function, is the gray domain Gaussian function, ||p - q|| is the spatial distance between pixel points, and ||I(p) - I(q)|| is the gray difference between pixel points.

[0039] Among them, after spatial domain noise reduction, frequency domain noise reduction is adopted to filter out salt-and-pepper noise, Gaussian noise, and Poisson noise in the night vision image. In frequency domain filtering, a band-pass filter is used for filtering. First, the night image is transformed from the spatial domain to the frequency domain by using the Fourier transform or wavelet transform method; then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range where the detailed information of the image is located; according to the analysis results, appropriate band-pass filter parameters are selected to design the band-pass filter.

[0040] Among them, the band-pass filter adopts Butterworth filter and Chebyshev filter, which have a smooth transition band and good filtering performance; the designed band-pass filter is applied to the image in the frequency domain to perform filtering processing on the image; the filtering processing includes direct filtering method or indirect filtering method. Among them, the direct filtering method multiplies the frequency response of the filter by the frequency spectrum of the image to obtain the filtered frequency spectrum, and then obtains the filtered spatial domain image through inverse transformation; the indirect filtering method first performs inverse transformation on the filter to obtain the filter in the spatial domain, then performs convolution operation on the filter and the image in the spatial domain to obtain the filtered image, and finally performs inverse transformation on the filtered frequency domain image to obtain the noise-reduced image in the spatial domain.

[0041] According to one or some embodiments of the present invention, the color low-light image generation module uses an interpolation algorithm to obtain a color low-light night vision image. First, the low-light image is converted from the RGB color space to the YUV color space, and the conversion formulas are respectively: Y = 0.299R + 0.587G + 0.114B; U = -0.147R - 0.289G + 0.436B; V = 0.615R - 0.515G - 0.100B, where R, G, and B respectively represent the red, green, and blue components in the RGB color space, and Y, U, and V respectively represent the luminance component and chrominance components in the YUV color space. Then, the values of unknown pixels are estimated by weighted averaging of adjacent pixels. During the process of converting a black-and-white image to a color image, the bilinear interpolation algorithm is used to estimate the missing color components. Among them, for each pixel point in the YUV image, the value of this pixel point is calculated by linearly weighting and averaging the values of its surrounding four known pixel points. Finally, the image in the processed YUV color space is converted back to the RGB color space to obtain a color low-light image.

[0042] According to another aspect of the present invention, a method for making a color low-light image is provided. The method for making a color low-light image is implemented based on a system for making a color low-light image. The system for making a color low-light image includes: a high-sensitivity CMOS imaging module, an image enhancement processing module, an image noise reduction processing module, and a color low-light image generation module.

[0043] The method for making a color low-light image includes the following steps:

[0044] S1. High-sensitivity CMOS imaging is achieved through the high-sensitivity CMOS imaging module to enhance the photon collection ability and reduce the noise level. The received light signal is converted into a charge signal by the built-in micro-photodiode array, and after being amplified, filtered, and digitized by the internal circuit, it is converted into digital image data to complete imaging.

[0045] S2. Image enhancement processing is performed on the imaged image. The histogram equalization algorithm is used, including calculating the gray histogram of the original image, calculating the cumulative distribution function according to the gray histogram, and performing gray mapping to obtain the enhanced gray value.

[0046] S3. Image noise reduction processing: Perform noise reduction on the enhanced image. Use night image noise reduction technology, and adopt bilateral filtering method including spatial domain noise reduction and band - pass filter for frequency domain noise reduction. Among them, the bilateral filtering method for spatial domain noise reduction combines spatial domain and gray - scale domain information to determine the filtering weight, and can better retain the edges and details of the image while removing noise; the design of the band - pass filter adopts the method of Butterworth filter or Chebyshev filter. First, convert the night image from the spatial domain to the frequency domain, determine the noise frequency range and the frequency range where the image detail information is located, design and apply the band - pass filter, and then perform inverse transformation to obtain the noise - reduced image in the spatial domain.

[0047] S4. Color low - light image generation: Use interpolation algorithm to obtain color low - light night vision images. Convert the low - illumination image from RGB space to YUV space, use bilinear interpolation algorithm to estimate the missing color components in the YUV image, calculate the value of each pixel point by linearly weighted averaging the values of four known pixel points around it, and finally convert the processed image in YUV color space back to RGB color space to obtain the color low - light image, completing the production of the color low - light image.

[0048] According to the system and method for making color low - light images of the present invention, the high - sensitivity CMOS sensor adopted has significantly enhanced photon collection ability and lower noise level compared with traditional sensors. This enables it to maintain high image clarity and signal - to - noise ratio even under extremely low - light conditions. This characteristic is particularly important for night, dim environments or specific application scenarios with limited light, which can significantly improve the imaging quality and capture more details that are difficult to detect by the naked eye.

[0049] According to one or some embodiments of the present invention, in step S1, the high - sensitivity CMOS image module pre - processes the image, including automatic exposure control (AEC) and automatic gain control (AGC), to dynamically adjust the exposure time and gain settings according to the change of ambient light, ensuring the best imaging effect under different light conditions. At the same time, through the built - in or external image - processing module, preliminary noise reduction and white - balance processing can be performed.

[0050] According to one or some embodiments of the present invention, in step S1, the image enhancement processing adopts a night - vision image enhancement algorithm for improving the quality of images obtained under low - light conditions to better observe and analyze the target. The night - vision image enhancement algorithm adopts histogram equalization algorithm for image enhancement processing, including the following steps:

[0051] First, calculate the grayscale histogram of the original image. Traverse each pixel of the original image, count the number of occurrences of each grayscale level, and obtain the grayscale histogram of the original image. For an image of size M×N with a grayscale range from 0 to L - 1, the grayscale histogram h(k) represents the number of occurrences of grayscale level k, where k = 0, 1, …, L - 1.

[0052] Secondly, calculate the cumulative distribution function c(k) according to the grayscale histogram. The formula is: where k = 0, 1, …, L - 1, and c(k) represents the total number of pixels with grayscale levels less than or equal to k.

[0053] Finally, perform grayscale mapping. Normalize the cumulative distribution function to obtain the mapped grayscale value. The normalization formula is: where M×N is the total number of pixels in the image and L is the total number of grayscale levels. For each pixel in the original image, according to its grayscale value k, obtain the enhanced grayscale value through the mapping relationship s(k).

[0054] According to one or some embodiments of the present invention, in step S3, for image noise reduction processing, a night image noise reduction technique is used to reduce the noise of the enhanced image, including spatial domain noise reduction, frequency domain noise reduction, and 3D noise reduction methods to reduce the noise of the enhanced image.

[0055] Among them, spatial domain noise reduction is a filtering method that operates on pixels in the image space. By performing weighted summation on each pixel in the image and its neighboring pixels, the value of the pixel is changed. The bilateral filtering method is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and grayscale domain information. By considering the spatial distance and grayscale difference between pixel points to determine the filtering weight, it can better retain the edges and details of the image while removing noise. The calculation formula for bilateral filtering is: where, I out (p) is the grayscale value of the output image at pixel point p, W p is the normalization factor, S is the neighborhood, is the spatial domain Gaussian function, is the grayscale domain Gaussian function, ||p - q|| is the spatial distance between pixel points, and ||I(p) - I(q)|| is the grayscale difference between pixel points.

[0056] Among them, after performing spatial domain noise reduction, frequency domain noise reduction is then adopted to filter out salt-and-pepper noise, Gaussian noise, and Poisson noise in the night vision image. In frequency domain filtering, a band-pass filter is used for filtering. First, the night image is transformed from the spatial domain to the frequency domain by using the method of Fourier transform or wavelet transform. Then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range where the detailed information of the image is located. According to the analysis results, appropriate band-pass filter parameters are selected to design the band-pass filter.

[0057] Among them, the band-pass filter adopts Butterworth filter and Chebyshev filter, which have a smooth transition band and good filtering performance. The designed band-pass filter is applied to the image in the frequency domain to perform filtering processing on the image. The filtering processing includes direct filtering method or indirect filtering method. Among them, in the direct filtering method, the frequency response of the filter is multiplied by the frequency spectrum of the image to obtain the filtered frequency spectrum, and then the filtered spatial domain image is obtained through inverse transformation. In the indirect filtering method, the filter is first inversely transformed to obtain the filter in the spatial domain, and then the filter is convolved with the image in the spatial domain to obtain the filtered image. Finally, the filtered frequency domain image is inversely transformed to obtain the noise-reduced image in the spatial domain.

[0058] According to one or some embodiments of the present invention, in step S4, the generation of the color low-light image adopts an interpolation algorithm to obtain the color low-light night vision image. First, the low-light image is transformed from the RGB space to the YUV space, and the transformation formulas are respectively: Y = 0.299R + 0.587G + 0.114B; U = -0.147R - 0.289G + 0.436B; 0.615R - 0.515G - 0.100B, where R, G, and B respectively represent the red, green, and blue components in the RGB color space, and Y, U, and V respectively represent the luminance component and chrominance components in the YUV color space. Then, the value of the unknown pixel is estimated by weighted averaging of adjacent pixels. In the process of converting a black-and-white image to a color image, the bilinear interpolation algorithm is used to estimate the missing color components. Among them, for each pixel point in the YUV image, the value of the pixel point is calculated by linearly weighting and averaging the values of its surrounding four known pixel points. Finally, the image in the processed YUV color space is converted back to the RGB color space to obtain the color low-light image.

[0059] According to the system and method for making a color low-light image of the present invention, in spatial domain noise reduction, the bilateral filtering method is adopted to determine the filtering weight by combining spatial domain and gray domain information, and the edges and detailed information of the image are better retained while removing noise.

[0060] The above description is only an exemplary implementation manner of the present invention, rather than used to limit the protection scope of the present invention. The protection scope of the present invention is determined by the appended claims.

Claims

1. A system for producing color low-light images, wherein: include: High-sensitivity CMOS imaging module, which is an image sensor used to enhance photon collection capability and reduce noise levels. It converts received light signals into charge signals through a built-in micro-photodiode array, which are then amplified, filtered and digitized by internal circuits to convert into digital image data to complete imaging; The image enhancement processing module is used to enhance the image by using a histogram equalization algorithm, including calculating the grayscale histogram of the original image, calculating the cumulative distribution function according to the grayscale histogram, and performing grayscale mapping to obtain the enhanced grayscale value; The image noise reduction processing module is used to reduce the noise of the enhanced image, adopt the night image noise reduction technology, and adopt the bilateral filtering method including spatial noise reduction and the bandpass filter of frequency domain noise reduction for filtering. Among them, the bilateral filtering method of spatial noise reduction combines the spatial domain and gray domain information to determine the filtering weight, and better retains the edge and detail information of the image while removing the noise; the bandpass filter design adopts the Butterworth filter or Chebyshev filter method, first converts the night image from the spatial domain to the frequency domain, determines the noise frequency range and the frequency range of the image detail information, designs and applies the bandpass filter, and then performs inverse transformation to obtain the denoised image in the spatial domain; A color low-light image generation module, wherein the color low-light image generation module adopts an interpolation algorithm to obtain a color low-light night vision image, converts the low-light image from RGB space to YUV space, estimates the missing color components in the YUV image using a bilinear interpolation algorithm, calculates the value of each pixel by taking a linear weighted average of the values ​​of four known pixels around each pixel, and finally converts the processed YUV color space image back to RGB color space to obtain a color low-light image, thereby completing the production of a color low-light image.

2. The system for producing a color low-light image as claimed in claim 1, wherein: The sensor of the high-sensitivity CMOS image module can maintain high image clarity and signal-to-noise ratio under extremely low light conditions, and is suitable for specific application scenarios at night, in dim environments or with limited light. The high-sensitivity CMOS image module pre-processes the image, including automatic exposure control AEC and automatic gain control AGC, to dynamically adjust the exposure time and gain setting according to changes in ambient light, ensuring that the best imaging effect can be obtained under different lighting conditions. At the same time, preliminary denoising and white balance processing can be performed through a built-in or external image processing module.

3. The system for producing color low-light images as claimed in claim 1, wherein: The image enhancement processing module adopts a night vision image enhancement algorithm to improve the image quality acquired under low light conditions so as to better observe and analyze the target. The night vision image enhancement algorithm adopts a histogram equalization algorithm to perform image enhancement processing, including the following steps: First, calculate the grayscale histogram of the original image. Traverse each pixel of the original image, count the number of occurrences of each gray level, and obtain the grayscale histogram of the original image. M×N The grayscale range of an image is from 0 to L-1, then the grayscale histogram h(k) represents the number of occurrences of grayscale k. Where, k = 0, 1, ... L-1; Secondly, the cumulative distribution function c(k) is calculated based on the grayscale histogram. The formula is: Wherein, k = 0, 1, ... L-1, c(k) represents the total number of pixels with gray level less than or equal to k; Finally, grayscale mapping is performed and the cumulative distribution function is normalized to obtain the mapped grayscale value; the normalization formula is: Where M×N is the total number of pixels in the image, and L is the total number of gray levels; for each pixel in the original image, according to its gray value k, the enhanced gray value is obtained through the mapping relationship s(k).

4. The system for producing a color low-light image as claimed in claim 1, wherein: The image noise reduction processing module uses night image noise reduction technology to reduce the noise of the enhanced image, including spatial domain noise reduction, frequency domain noise reduction and 3D noise reduction methods to reduce the noise of the enhanced image. Among them, spatial domain denoising is a filtering method that operates on pixels in the image space. The pixel value is changed by weighted summing of each pixel in the image and its neighboring pixels. Bilateral filtering is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and grayscale domain information. The filtering weight is determined by considering the spatial distance and grayscale difference between pixel points. It can better retain the edge and detail information of the image while removing noise. The calculation formula of bilateral filtering is: Among them, I out (p) is the grayscale value of the output image at pixel p, W p is the normalization factor, S is the neighborhood, is a Gaussian function in the spatial domain, is the grayscale domain Gaussian function, ||pq|| is the spatial distance between pixels and , ||I(p)-I(q)|| is the grayscale difference between pixels and ; Among them, after performing spatial domain denoising, frequency domain denoising is used to filter out salt and pepper noise, Gaussian noise and Poisson noise in night vision images. In frequency domain filtering, bandpass filters are used for filtering. First, the night image is converted from the spatial domain to the frequency domain by using Fourier transform or wavelet transform methods; then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range of the image detail information; according to the analysis results, appropriate bandpass filter parameters are selected to design the bandpass filter; Among them, the bandpass filter adopts Butterworth filter and Chebyshev filter, which have smooth transition band and good filtering performance; the designed bandpass filter is applied to the image in the frequency domain to filter the image; the filtering process includes direct filtering method or indirect filtering method, wherein the direct filtering method multiplies the frequency response of the filter with the frequency spectrum of the image to obtain the filtered frequency spectrum, and then obtains the filtered spatial domain image through inverse transformation; the indirect filtering method first inverse transforms the filter to obtain the filter in the spatial domain, and then convolves the filter with the image in the spatial domain to obtain the filtered image, and finally, inverse transforms the filtered frequency domain image to obtain the denoised image in the spatial domain.

5. The system for producing color low-light images as claimed in claim 1, wherein: The color low-light image generation module adopts an interpolation algorithm to obtain a color low-light night vision image. First, the low-light image is converted from the RGB space to the YUV space. The conversion formulas are: Y=0.299R+0.587G+0.114BU=-0.147R-0.289G+0.436B.0.615R-0.515G-0.100B, wherein R, G, and B represent the red, green, and blue components in the RGB color space, respectively, and Y, U, and V represent the brightness component and the chromaticity component in the YUV color space, respectively; then, the value of the unknown pixel is estimated by weighted averaging the adjacent pixels. In the process of converting the black-and-white image to the color image, the missing color component is estimated by using a bilinear interpolation algorithm; wherein, for each pixel in the YUV image, the interpolation method calculates the value of the pixel by performing a linear weighted average of the values ​​of the four known pixels around it; finally, the processed image in the YUV color space is converted back to the RGB color space to obtain a color low-light image.

6. A method for producing a color low-light image, wherein the method is implemented based on a system for producing a color low-light image, and the system for producing a color low-light image comprises: High-sensitivity CMOS imaging module, image enhancement processing module, image noise reduction processing module and color low-light image generation module; The method for producing a color low-light image comprises the following steps: S1. High-sensitivity CMOS imaging is achieved through the high-sensitivity CMOS imaging module, which can enhance the photon collection capability and reduce the noise level. The received light signal is converted into a charge signal through the built-in micro-photodiode array, and then amplified, filtered and digitized by the internal circuit to convert it into digital image data to complete the imaging; S2, image enhancement processing, performing enhancement processing on the imaged image, using a histogram equalization algorithm, including calculating a grayscale histogram of the original image, calculating a cumulative distribution function according to the grayscale histogram, and performing grayscale mapping to obtain an enhanced grayscale value; S3, image denoising, denoising the enhanced image, using night image denoising technology, filtering using bilateral filtering method including spatial denoising and bandpass filter for frequency domain denoising, wherein the bilateral filtering method for spatial denoising combines spatial domain and grayscale domain information to determine the filtering weight, while removing noise, better retaining the edge and detail information of the image; the bandpass filter design adopts the Butterworth filter or Chebyshev filter method, first converting the night image from the spatial domain to the frequency domain, determining the noise frequency range and the frequency range of the image detail information, designing and applying the bandpass filter, and then performing inverse transformation to obtain the denoised image in the spatial domain; S4. Color low-light image generation: an interpolation algorithm is used to obtain a color low-light night vision image. The low-light image is converted from RGB space to YUV space. A bilinear interpolation algorithm is used to estimate the missing color components in the YUV image. The value of each pixel is calculated by taking a linear weighted average of the values ​​of the four known pixels around each pixel. Finally, the processed YUV color space image is converted back to RGB color space to obtain a color low-light image, thus completing the production of a color low-light image.

7. The method for producing a color low-light image according to claim 6, wherein: In step S1, the high-sensitivity CMOS image module pre-processes the image, including automatic exposure control AEC and automatic gain control AGC, to dynamically adjust the exposure time and gain setting according to changes in ambient light, to ensure the best imaging effect under different lighting conditions, and at the same time, preliminary denoising and white balance processing can be performed through the built-in or external image processing module.

8. The method for producing a color low-light image according to claim 6, wherein: In step S1, the image enhancement process uses a night vision image enhancement algorithm to improve the image quality acquired under low light conditions so as to better observe and analyze the target. The night vision image enhancement algorithm uses a histogram equalization algorithm to perform image enhancement process, including the following steps: First, calculate the grayscale histogram of the original image. Traverse each pixel of the original image, count the number of occurrences of each gray level, and obtain the grayscale histogram of the original image. M×N The grayscale range of an image is from 0 to L-1, then the grayscale histogram h(k) represents the number of occurrences of grayscale k. Where, k = 0, 1, ... L-1; Secondly, the cumulative distribution function c(k) is calculated based on the grayscale histogram. The formula is: Wherein, k = 0, 1, ... L-1, c(k) represents the total number of pixels with gray level less than or equal to k; Finally, grayscale mapping is performed and the cumulative distribution function is normalized to obtain the mapped grayscale value; the normalization formula is: Where M×N is the total number of pixels in the image, and L is the total number of gray levels; for each pixel in the original image, according to its gray value k, the enhanced gray value is obtained through the mapping relationship s(k).

9. The method for producing a color low-light image according to claim 6, wherein: In step S3, the image denoising process uses night image denoising technology to denoise the enhanced image, including spatial domain denoising, frequency domain denoising and 3D denoising methods to denoise the enhanced image. Among them, spatial domain denoising is a filtering method that operates on pixels in the image space. The pixel value is changed by weighted summing of each pixel in the image and its neighboring pixels. Bilateral filtering is used for spatial domain filtering. The bilateral filtering method is a filtering method that combines spatial domain and grayscale domain information. The filtering weight is determined by considering the spatial distance and grayscale difference between pixel points. It can better retain the edge and detail information of the image while removing noise. The calculation formula of bilateral filtering is: Among them, I out (p) is the grayscale value of the output image at pixel p, W p is the normalization factor, S is the neighborhood, is a Gaussian function in the spatial domain, is the grayscale domain Gaussian function, ||pq|| is the spatial distance between pixels and , ||I(p)-I(q)|| is the grayscale difference between pixels and ; Among them, after performing spatial domain denoising, frequency domain denoising is used to filter out salt and pepper noise, Gaussian noise and Poisson noise in night vision images. In frequency domain filtering, bandpass filters are used for filtering. First, the night image is converted from the spatial domain to the frequency domain by using Fourier transform or wavelet transform methods; then, the image in the frequency domain is analyzed to determine the frequency range of the noise and the frequency range of the image detail information; according to the analysis results, appropriate bandpass filter parameters are selected to design the bandpass filter; Among them, the bandpass filter adopts Butterworth filter and Chebyshev filter, which have smooth transition band and good filtering performance; the designed bandpass filter is applied to the image in the frequency domain to filter the image; the filtering process includes direct filtering method or indirect filtering method, wherein the direct filtering method multiplies the frequency response of the filter with the frequency spectrum of the image to obtain the filtered frequency spectrum, and then obtains the filtered spatial domain image through inverse transformation; the indirect filtering method first inverse transforms the filter to obtain the filter in the spatial domain, and then convolves the filter with the image in the spatial domain to obtain the filtered image, and finally, inverse transforms the filtered frequency domain image to obtain the denoised image in the spatial domain.

10. The method for producing a color low-light image according to claim 6, wherein: In step S4, the color low-light image is generated by using an interpolation algorithm to obtain a color low-light night vision image. First, the low-light image is converted from the RGB space to the YUV space. The conversion formulas are: Y = 0.299R + 0.587G + 0.114B; U = -0.147R - 0.289G + 0.436B; 0.615R-0.515G-0.100B, where R, G, and B represent the red, green, and blue components in the RGB color space, respectively, and Y, U, and V represent the brightness component and chrominance component in the YUV color space, respectively; then, the value of the unknown pixel is estimated by taking the weighted average of adjacent pixels, and in the process of converting the black-and-white image to a color image, the missing color component is estimated by using the bilinear interpolation algorithm; wherein, for each pixel in the YUV image, the interpolation method calculates the value of the pixel by taking the linear weighted average of the values ​​of the four surrounding known pixels; finally, the processed image in the YUV color space is converted back to the RGB color space to obtain a color low-light image.

Images