A non-uniform adaptive image enhancement method based on low illumination

Abstract

The invention discloses a non-uniform adaptive image enhancement method based on low illumination, comprising the following steps: S1. converting a color image space into a YCbCr space to obtain an original brightness Y component of an input image; S2. performing guided filtering on the original brightness Y component to enhance image details; S3. performing iterative calculations through a cuckoo search algorithm to obtain the optimal parameters of a sigmoid function when the image contrast is maximized, and obtaining an optimized sigmoid function; S4. using the optimized sigmoid function to linearly enhance the brightness image after detail enhancement, and obtaining an enhanced brightness Y component; S5. constructing a color scale factor according to the ratio of the enhanced brightness Y component to the original brightness Y component, and obtaining an enhanced visible light image through the color scale factor and R, G and B components of the original image.

Description

Technical Field

[0001] The present invention relates to the technical field of digital image processing, and more particularly to a non-uniform adaptive image enhancement method under low illumination. Background Art

[0002] With the rapid development of digital imaging technology, various types of imaging optical systems are being used more and more widely. However, in the evening, at night or in poorly lit indoor environments, the images acquired by imaging devices usually have low brightness and contrast, and serious loss of image details, which seriously affects the performance of tasks such as target detection and recognition. How to efficiently and quickly enhance low-light images is an important issue that needs to be solved in the field of image processing.

[0003] Therefore, how to propose a non-uniform adaptive image enhancement method under low illumination is a problem that those skilled in the art need to solve urgently. Summary of the invention

[0004] In view of this, the present invention provides a non-uniform adaptive image enhancement method based on low illumination, which is used to solve the problem of halo artifacts and insufficient enhancement details in image enhancement while effectively enhancing non-uniform low-light images.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] A non-uniform adaptive image enhancement method under low illumination comprises the following steps:

[0007] S1. Convert the color image space to the YCbCr space to obtain the original brightness Y component of the input image;

[0008] S2. Performing guided filtering on the original brightness Y component to enhance image details;

[0009] S3. Iteratively calculate the optimal parameters of the sigmoid function when the image contrast is maximized by using the cuckoo search algorithm to obtain the optimized sigmoid function;

[0010] S4. Use the optimized sigmoid function to linearly enhance the detail-enhanced brightness image to obtain an enhanced brightness Y component;

[0011] S5. Construct a color scale factor according to the ratio of the enhanced brightness Y component to the original brightness Y component, and obtain an enhanced visible light image through the color scale factor and the R, G and B components of the original image.

[0012] Preferably, the specific content of performing guided filtering on the original brightness Y component to enhance image details in S2 includes:

[0013]

[0014] Where i is the window ω k The pixel position in , q is the guided filter output image, I is the guided image, where a k and b k For window w k The linear coefficients in are:

[0015]

[0016]

[0017] Where p is the input image, w k is a window centered at pixel k and is defined as:

[0018]

[0019] Among them, |ω| is ω k The total number of pixels in the and μ k They are ω k The variance and mean of ; is the ω in image p k The mean value of pixels within; ε is the regularization factor;

[0020] Combined linear coefficient a k ,b k , using different window sizes to obtain different guided filter output images q:

[0021]

[0022] When the guided image I is consistent with the input image P, the guided filter is an edge-preserving filter, and the brightness Y component of the low-illuminance input image is subjected to self-guided filtering to obtain a guided filter map of the filtered brightness Y component. The filtered brightness Y component is subtracted from the original brightness Y component to obtain the details of the image, and the details of the image are amplified and added to the original brightness Y component to obtain an image with enhanced details.

[0023] Preferably, the specific content of the optimized sigmoid function in S3 includes:

[0024] Establish the improved sigmoid function:

[0025]

[0026] a, b are the coefficients to be determined, and the image contrast C is enhanced by calculating the sigmoid functiona , with C a The maximum a and b are the optimal values ​​of the sigmoid function, and the image contrast is calculated as follows:

[0027]

[0028] Where δ(i,j)=|ij| is the grayscale difference between adjacent pixels, P δ (i, j) is the pixel distribution probability when the grayscale difference between adjacent pixels is δ.

[0029] Preferably, C is calculated by a cuckoo search algorithm a The values ​​of a and b are the maximum, and the specific contents of the cuckoo search algorithm include:

[0030] (1) Randomly generate N bird nests, i.e., a and b values, select an optimal bird nest location in the current era, and keep it in the next iteration;

[0031] (2) Update the Levy flight position to obtain the next generation position in each bird's nest, compare it with the previous generation position, and select the one that makes the image contrast C a The largest a and b values ​​replace the current a and b values;

[0032] (3) The probability of foreign bird eggs P a Compared with the random number r∈[0,1]. When r>P a , the position of the bird's nest can be changed, otherwise, the original position is maintained, and then the image contrast C is compared. a , select C a The a and b values ​​at maximum;

[0033] (4) When the bird's nest reaches the accuracy requirement or the number of iterations reaches the maximum value, the a and b values ​​are set as the global optimal solution, otherwise continue to execute (2).

[0034] It can be seen from the above technical solution that compared with the prior art, the present invention discloses a non-uniform adaptive image enhancement method based on low illumination. Compared with the existing algorithm, the algorithm proposed in the present invention can effectively enhance the non-uniform low-light image while improving the problems of halo artifacts and insufficient enhancement details that are prone to occur in image enhancement, which can greatly improve the efficiency of subsequent target detection, recognition and other tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0036] Figure 1 The accompanying figure is a flow chart of a non-uniform image adaptive enhancement method under low illumination provided by the present invention. Where img 1 is the RGB color space of the original image, and the Y brightness component, Cb blue component and Cr red component shown in the figure are obtained by converting RGB→YCbCr; img 2 is obtained by guided filtering of the Y component; img 3 is the Y component enhanced by the optimized sigmoid function; the right box diagram is a flow chart of the CS algorithm seeking the optimal sigmoid parameter; a scale factor is constructed by img 3 and the original Y component, and img 4 is the enhanced RGB component obtained by multiplying img 1 by the scale factor; img 5 is the enhanced original image.

[0037] Figure 2 The accompanying drawing is a schematic diagram of a curve of a sigmoid function provided in an embodiment of the present invention;

[0038] Figure 3 The accompanying drawing is a schematic diagram of experimental results provided by an embodiment of the present invention, wherein the left column is the original low-light image, and the right column is the corresponding enhanced low-light image. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] The embodiment of the present invention discloses a non-uniform adaptive image enhancement method under low illumination, such as Figure 1 As shown, the following steps are included:

[0041] S1. Convert the color image space to the YCbCr space to obtain the original brightness Y component of the input image;

[0042] S2. Perform guided filtering on the original brightness Y component to enhance image details;

[0043] S3. Iteratively calculate the optimal parameters of the sigmoid function when the image contrast is maximized by using the cuckoo search (CS) algorithm to obtain the optimized sigmoid function;

[0044] S4. Use the optimized sigmoid function to linearly enhance the detail-enhanced brightness image to obtain an enhanced brightness Y component;

[0045] S5. Construct a color scale factor according to the ratio of the enhanced brightness Y component to the original brightness Y component, and obtain an enhanced visible light image through the color scale factor and the R, G and B components of the original image.

[0046] In order to further implement the above technical solution, the specific contents of performing guided filtering on the original brightness Y component to enhance the image details in S2 include:

[0047]

[0048] Where i is the window ω k The pixel position in , q is the guided filter output image, I is the guided image, where a k and b k For window w k The linear coefficients in are:

[0049]

[0050]

[0051] Where p is the input image, w k is a window centered at pixel k and is defined as:

[0052]

[0053] Among them, |ω| is ω k The total number of pixels in the and μ k They are ω k The variance and mean of ; is the ω in image p k The mean value of pixels within; ε is the regularization factor;

[0054] Combined linear coefficient a k ,b k , using different window sizes to obtain different guided filter output images q:

[0055]

[0056] When the guided image I is consistent with the input image P, the guided filter becomes an edge-preserving filter. Using this property, the brightness Y component of the input low-light image is self-guided filtered to obtain the guided filter map of the brightness Y component. The guided filter map is subtracted from the brightness Y component of the original image to obtain the details of the image. The obtained image details are amplified and added to the brightness Y component of the original image, which makes the details of the image more profound, making up for the problem of insufficient enhanced image details in traditional algorithms.

[0057] It should be noted that:

[0058] Guided filtering defines a linear filtering process, using a guide image as the filter content image to implement a local linear function expression on the input image. Guided filtering is used as a local linear model to describe the relationship between the guide image I and the output image q.

[0059] The normalization factor ε is used to prevent the coefficient a k Too large, it is an important parameter for adjusting the filter.

[0060] In order to further implement the above technical solution, the specific contents of the optimized sigmoid function in S3 include:

[0061] Establish the improved sigmoid function:

[0062]

[0063] a, b are the coefficients to be determined, and the image contrast C is enhanced by calculating the sigmoid function a , with C a The maximum a and b are the optimal values ​​of the sigmoid function, and the image contrast is calculated as follows:

[0064]

[0065] Where δ(i,j)=|ij| is the grayscale difference between adjacent pixels, P δ (i, j) is the pixel distribution probability when the grayscale difference between adjacent pixels is δ.

[0066] In order to further implement the above technical solution, C is calculated by the cuckoo search algorithm. a The values ​​of a and b are the maximum. The specific contents of the Cuckoo Search (CS) algorithm include:

[0067] (1) Randomly generate N bird nests, i.e., a and b values, select an optimal bird nest location in the current era, and keep it in the next iteration;

[0068] (2) Update the Levy flight position to obtain the next generation position in each bird's nest, compare it with the previous generation position, and select the one that makes the image contrast C a The largest a and b values ​​replace the current a and b values;

[0069] (3) The probability of foreign bird eggs P a Compared with the random number r∈[0,1]. When r>P a , the position of the bird's nest can be changed, otherwise, the original position is maintained, and then the image contrast C is compared. a , select C a The a and b values ​​at maximum;

[0070] (4) When the bird's nest reaches the accuracy requirement or the number of iterations reaches the maximum value, the a and b values ​​are set as the global optimal solution, otherwise continue to execute (2).

[0071] It should be noted that:

[0072] The Sigmoid function can map variables to between [0, 1], such as Figure 2 As shown in the figure, the sigmoid function can better enhance the dark area while suppressing the brightness of the bright area, which can make up for the fact that the gamma function cannot take into account both the enhancement of the dark area and the suppression of the bright area at the same time. The determination of the function value is transformed into an optimization problem, that is, selecting the values ​​of a and b when the contrast is optimal to adaptively construct the enhancement function.

[0073] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0074] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A non-uniform adaptive image enhancement method based on low illumination, characterized in that: The following steps are involved: S1. Convert the color image space to the YCbCr space to obtain the original brightness Y component of the input image; S2. Performing guided filtering on the original brightness Y component to enhance image details; S3. Iteratively calculate the optimal parameters of the sigmoid function when the image contrast is maximized by using the cuckoo search algorithm to obtain the optimized sigmoid function; S4. Use the optimized sigmoid function to linearly enhance the detail-enhanced brightness image to obtain an enhanced brightness Y component; S5. constructing a color scale factor according to the ratio of the enhanced brightness Y component to the original brightness Y component, and obtaining an enhanced visible light image through the color scale factor and the R, G and B components of the original image; The specific contents of performing guided filtering on the original brightness Y component to enhance the image details in S2 include: Where i is the window ω k The pixel position in , q is the guided filter output image, I is the guided image, where a k and b k For window ω k The linear coefficients in are: Where p is the input image, ω k is a window centered at pixel k and is defined as: Among them, |ω| is ω k The total number of pixels in the and μ k They are ω k The variance and mean of ; is the ω in image p k The pixel mean within; ε is the regularization factor; Combined linear coefficient a k ,b k , using different window sizes to obtain different guided filter output images q: When the guide image I is consistent with the input image P, the guide filter is an edge-preserving filter, and the brightness Y component of the low-illuminance input image is subjected to self-guided filtering to obtain a guided filter map of the filtered brightness Y component, and the filtered brightness Y component is subtracted from the original brightness Y component to obtain the detail part of the image, and the detail part of the image is amplified and added to the original brightness Y component to obtain an image with enhanced details; The specific contents of the optimized sigmoid function in S3 include: Establish the improved sigmoid function: a, b are the coefficients to be determined, and the image contrast C is enhanced by calculating the sigmoid function a , with C a The maximum a and b are the optimal values ​​of the sigmoid function, and the image contrast is calculated as follows: Where δ(i,j)=|ij| is the grayscale difference between adjacent pixels, P δ (i, j) is the pixel distribution probability when the grayscale difference between adjacent pixels is δ.

2. The non-uniform adaptive image enhancement method under low illumination according to claim 1, characterized in that: Calculate C through the cuckoo search algorithm a The values ​​of a and b are the maximum, and the specific contents of the cuckoo search algorithm include: (1) Randomly generate N bird nests, i.e., a and b values, select an optimal bird nest location in the current era, and keep it in the next iteration; (2) Update the Levy flight position to obtain the next generation position in each bird's nest, compare it with the previous generation position, and select the one that makes the image contrast C a The largest a and b values ​​replace the current a and b values; (3) The probability of foreign bird eggs P a Compared with the random number r∈[0,1]; when r>P a , then change the position of the bird's nest, otherwise, keep the original position, and then compare the image contrast C a , select C a The a and b values ​​at maximum; (4) When the bird's nest reaches the accuracy requirement or the number of iterations reaches the maximum value, the a and b values ​​are set as the global optimal solution, otherwise continue to execute (2).

Images