An uncertain illumination image enhancement method and system

Abstract

The present application provides a kind of uncertain illumination image enhancement method, comprising: using guide filter method to decompose the image to be enhanced into base layer and detail layer;According to the brightness of base layer, the detail layer is stretched to obtain an enhanced detail layer, and then the enhanced detail layer and the base layer are fused to obtain an enhanced image;Using the adaptive brightness adjustment method based on image entropy to adjust the enhanced image to obtain the final image.The technical scheme of the present application solves the problem of detail loss in low-light area and over-illuminated area during the enhancement process of the existing enhancement algorithm, and has lower computational cost while improving the quality of the enhanced image.

Description

Technical Field

[0001] This invention pertains to the field of image processing methods, specifically relating to a guided filtering and adaptive intensity enhancement method and system for image enhancement under uncertain illumination. Background Technology

[0002] In natural environments, illumination is typically complex, variable, and uncertain. The uncertainty of ambient lighting leads to drastic changes in the dynamic range of light intensity within a scene, while the dynamic range of image sensors is limited. Improper exposure can result in poor image contrast. This is particularly true in low-light or over-lit areas, where numerous details remain undisplayed, degrading image quality. In many practical applications, such as surveillance, remote sensing imaging, and computer vision, we need to improve the contrast of such uncertainly illuminated images to enhance image quality, providing more suitable data for further processing and ultimately improving processing performance in these fields. Therefore, research on enhancing uncertainly illuminated images has significant theoretical and practical value.

[0003] In recent years, research on image enhancement under uncertain lighting has received considerable attention, including histogram equalization methods and their improvements, Retinex-based methods and their improvements, and deep learning-based methods. However, these methods each have their own limitations in terms of their underlying principles. For example, histogram-based methods process each pixel differently based on its features, effectively increasing local contrast, but they also have limitations such as long processing times and inability to adapt to backlighting. Retinex-based methods focus on effectively estimating the weights of background illumination. By directly subtracting the background illumination component from the original image in logarithmic space, they enhance the overall image contrast but fail to improve local details. Deep learning-based algorithms can improve image quality in multiple ways, but require substantial computational resources. Summary of the Invention

[0004] Based on research into existing methods and addressing their existing problems, this invention provides a method for enhancing images with uncertain illumination, comprising:

[0005] S1: The guided filtering method is used to decompose the image to be enhanced into a base layer and a detail layer;

[0006] S2: Stretch the detail layer according to the brightness of the base layer to obtain an enhanced detail layer, and then blend the enhanced detail layer with the base layer to obtain the enhanced image;

[0007] S3: The enhanced image is adjusted using an adaptive brightness adjustment method based on image entropy to obtain the final image.

[0008] Furthermore, in step S1, the decomposition process is represented as follows:

[0009] I base =G f (I in )

[0010] I detail =I in -I base

[0011] I in For the image to be enhanced, G f Let I be the guided filter transform function, where I detail ,I base These represent the detail layer and the base layer obtained after guided filter decomposition, respectively.

[0012] Furthermore, in step S2, the stretching enhancement function that stretches the detail layer image based on the base layer that can characterize the background light intensity is:

[0013]

[0014] Among them I detial ,I base This represents the detail layer and the base layer obtained after guided filtering decomposition, I detial_p This represents the detail layer after stretching and enhancement.

[0015] Furthermore, in step S2, the fusion process is represented as follows:

[0016] I enhance =I base +η*I detial_p

[0017] Among them I base I represents the base layer obtained after guided filtering decomposition. detial_p I represents the detail layer after stretching transformation. enhance The image represents the enhanced image, and η represents the detail layer I after stretching and enhancement during the fusion process. detial_p The fusion coefficient.

[0018] Furthermore, in S3, the adaptive brightness adjustment method based on image entropy specifically includes the following steps:

[0019] S301: Output image I after adaptive brightness adjustment out The dispersion of brightness, i.e. image entropy, determines the gamma transform parameter β that maximizes the entropy of the output image;

[0020] S302: Based on the enhanced image I enhanceThe logarithmic mean of the illumination intensity is used to determine the adaptive parameter x' used in the classification and transformation process;

[0021] S303: Based on the enhanced image I enhance Brightness is determined by dividing all pixels into high-brightness pixels and low-brightness pixels using the adaptive parameter x' as the dividing point.

[0022] S304: Perform brightness transformation on the two types of pixels in step S303 according to the determined gamma transformation parameter β and adaptive parameter x' respectively.

[0023] Furthermore, in S303, the enhanced image I enhance The pixel division is described as follows:

[0024]

[0025] Where x' is the adaptive parameter, L p (x,y) represents the gray levels of the pixels in the enhanced image, L M1 ,L M2 I is the transformation function for pixels of different gray levels. out (x,y) represents the output image pixels after pixel transformation;

[0026] The transformation function of the pixels after partitioning is expressed as follows:

[0027]

[0028]

[0029] Where x' is the adaptive parameter, L p (x,y) represents the gray levels of the pixels in the enhanced image, L M1 (x,y),L M2 (x,y) represents the enhanced image I. enhance The results of pixel transformation at different gray levels.

[0030] Furthermore, in step S302, an adaptive parameter x' is designed to enhance the image I. enhance When the overall illumination is low, the area with even lower illumination is primarily enhanced, resulting in an enhanced image I. enhance When the overall illuminance is high, the area with even higher illuminance should be enhanced.

[0031] The adaptive parameters are designed as follows:

[0032]

[0033] Where x' is the adaptive parameter, N is the number of pixels in the input image, δ is a very small positive number to avoid the logarithmic independent variable being 0, and L p(x,y) represents the enhanced image I. enhance pixel grayscale

[0034] According to another aspect of the present invention, an uncertain illumination image enhancement system is also provided, including a guided filtering layering module for decomposing the image to be enhanced into a base layer and a detail layer;

[0035] The detail layer enhancement adjustment module is used to stretch the detail layer based on the brightness of the base layer to obtain an enhanced detail layer; then the stretched and enhanced detail layer is blended with the base layer.

[0036] The adaptive brightness adjustment module uses an image entropy-based adaptive brightness adjustment method to adjust the enhanced image to obtain the final image.

[0037] The technical solution of this invention utilizes guided filtering to layer the image, effectively preserving edge information while layering. Simultaneously, by adjusting the detail layer and the overall brightness of the image, it achieves high-quality enhancement of images under uncertain lighting conditions. Therefore, the technical solution of this invention solves the problems of low enhancement quality and high computational resource requirements in existing technologies. Based on the above reasons, this invention can be widely applied in the field of image enhancement for uncertain images. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of the implementation method of the present invention;

[0040] Figure 2 This is a diagram illustrating the implementation method of the guided filtering image layering module of the present invention;

[0041] Figure 3 This is a diagram showing the effect of the arctangent function stretching of the detail layer stretching module in this invention.

[0042] Figure 4 This is a diagram illustrating the implementation steps of the adaptive brightness adjustment module of the present invention;

[0043] Figure 5 This is the adaptive brightness adjustment curve of the present invention; Detailed Implementation

[0044] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0045] The specific implementation of the present invention will now be described in detail with reference to the accompanying drawings.

[0046] like Figure 1-5 As shown, this invention proposes a method for enhancing images with uncertain illumination, comprising:

[0047] S1: The guided filtering method is used to decompose the image to be enhanced into a base layer and a detail layer;

[0048] More figures to be enhanced (Figure I) in Based on edge information, it is decomposed into base layer I. base and detail layer I detail The guided filtering method is used to enhance the image I. in Decomposed into base layer I base and detail layer I detail .

[0049] The decomposition process is described as follows:

[0050] I base =G f (I in )

[0051] I detail =I in -I base

[0052] I in For the image to be enhanced, G f Let I be the guided filter transform function, where I detail ,I base These represent the detail layer and the base layer obtained from the guided filtering-based image layering module, respectively.

[0053] Guided filtering is used to enhance the image, dividing it into a base layer and a detail layer. The base layer includes not only smooth areas but also particularly sharp edges, while subtle details are separated into detail layers. Post-processing using an edge-preserving filter allows the base layer to maintain its basic shape, while the detail layers extract texture for further enhancement. Compared to other filtering layering methods, guided filtering-based image layering methods better preserve edge information in the source input image.

[0054] S2: Reference Figure 3 The detail layer is stretched based on the brightness of the base layer to obtain an enhanced detail layer. The enhanced detail layer is then blended with the base layer to obtain the enhanced image.

[0055] Weber's Law states that the human eye can perceive minute differences between adjacent pixels in dark areas, while in bright areas, this difference must be large enough to be perceived. The detail layer enhancement adjustment module uses this principle to stretch the detail layer I of the image layering module based on guided filtering. detail When in the image I to be enhanced in In low-light areas, detail layer I detail It can be stretched less when in the image to be enhanced I in When the bright area is visible, detail layer I detail It can stretch further, that is, it uses the arctangent function to enhance the detail layer to different degrees. This detail stretching method is particularly effective for weak detail enhancement, while very obvious details are hardly enhanced.

[0056] The base layer I obtained after the guided filtering-based image layering module is... base It can characterize the image I to be enhanced in The light intensity, therefore, the stretching function of the detail layer is expressed as:

[0057]

[0058] Among them I detial ,I base This represents the detail layer and the base layer obtained after guided filtering decomposition, I detial_p This represents the detail layer after stretching and enhancement.

[0059] After stretching the detail layer, it is necessary to enhance the detail layer I. detail_p and base layer I base Re-fusion yields the enhanced image I enhance To avoid detail layer I in low-light areas detail The magnification is too low, which is the base layer I during blending. base A "minimum value" was set, which equals 0.3. This means that when blending, the base layer I... base When the actual value is less than 0.3, it will be set to 0.3. Furthermore, due to the stretching of the arctangent function, some details are greatly enhanced. When these details are in overly bright areas, detail layer I... detailThe components will become very large, so the stretching factor should not be too large; here, we use 0.8. Detail layer I after stretching and enhancement. detail_p With base layer diagram I base The fusion between them can be described as follows:

[0060] I enhance =I base +η*I detial_p

[0061] Among them I base I represents the base layer obtained by the guided filtering-based image layering module for enhancing the image. detial_p This represents the detail layer after stretching and enhancement processing obtained through the detail layer enhancement adjustment module. enhance This represents the enhanced image obtained after the detail layer enhancement adjustment module, and η represents the fusion coefficient of the detail layer after stretching enhancement during fusion.

[0062] S3: An adaptive brightness adjustment method based on image entropy is used to adjust the enhanced image to obtain the final image. (Reference) Figure 4-5 Adaptive brightness adjustment specifically includes the following steps:

[0063] S301: Output image I after adaptive brightness adjustment out The dispersion of brightness, i.e. image entropy, determines the gamma transform parameter β that maximizes the entropy of the output image;

[0064] As β increases, low-light and over-lit areas can achieve stronger enhancement, but this can also easily lead to over-enhancement, resulting in excessively low contrast in normally lit areas.

[0065] The "entropy" of an image represents the degree of dispersion in the image's brightness distribution, and can reflect the overall contrast of the image. To make the output image I... out The overall contrast effect is the best. Therefore, we determine β here guided by maximum entropy. In the implementation process, we select β values ​​of 1, 2, 3, 4, and 8 for the enhanced image I. enhance The output image I is obtained by adaptive brightness adjustment. out When β equals 3, the output image I out The entropy is the largest, so the value of β is chosen to be 3.

[0066] S302: Based on the enhanced image I enhance The logarithmic mean of the illumination intensity is used to determine the adaptive parameter x' used in the classification and transformation process;

[0067] x' is the dividing point between "low illuminance" and "high illuminance," and is also an adaptive parameter. It is affected by image brightness attribution and is related to the logarithmic mean of illumination intensity, with the following expression:

[0068]

[0069] Where x' is the adaptive parameter, N is the number of pixels in the input image, δ is a very small positive number, which can take the value 0.1 to avoid the logarithmic independent variable being 0, and L p (x,y) represents the enhanced image I. enhance pixel grayscale

[0070] When the enhanced image I enhance When the illumination is low, the logarithmic mean of the brightness is also low, making it desirable to enhance low-illuminance areas. In this case, a larger x' is desirable. When the enhanced image I... enhance When the illuminance is high, the adaptive parameter x' can be used to primarily enhance the high illuminance area.

[0071] S303: Based on the enhanced image I enhance Brightness is determined by dividing all pixels into high-brightness pixels and low-brightness pixels using the adaptive parameter x' as the dividing point.

[0072] Image illuminance typically has a large dynamic range, and insufficient contrast in underlit and overlit areas can make image details difficult to see clearly. Experiments have shown that the stimulation of the retina by illumination is directly proportional to a power function of illuminance; in other words, the response of retinal nerves is directly proportional to the gamma transformation of physical illumination, as shown below:

[0073]

[0074] L p =0.299*R + 0.587*G + 0.114*B

[0075] Where L p L represents the original strength. M The intensity is represented by W and γ, which are constants. The goal of enhancement is to increase the contrast in low-illumination areas while decreasing the contrast in high-illumination areas. For the image to be clearly visible, the difference in high-illumination areas must be sufficiently large. Therefore, in step 1, based on the enhanced image I... enhance Brightness is divided into two categories, and then two different methods are used to adjust the brightness, as shown below:

[0076] Enhanced image I enhance The pixel division is described as follows:

[0077]

[0078] Where x' is the adaptive parameter described in step 3, L p (x,y) represents the enhanced image I.enhance Pixel grayscale level, L M1 ,L M2 I is the transformation function for pixels of different gray levels. out (x,y) represents the output image pixels after pixel transformation.

[0079] The transformation function after pixel partitioning is expressed as follows:

[0080]

[0081]

[0082] Where x' is the adaptive parameter, L p (x,y) represents the enhanced image I. enhance Pixel grayscale level, L M1 (x,y),L M2 (x,y) represents the enhanced image I. enhance The results of pixel transformation at different gray levels.

[0083] S304: Perform brightness transformation on the two types of pixels in step S303 according to the determined gamma transformation parameter β and adaptive parameter x' respectively.

[0084] The adaptive brightness adjustment method selects parameters that maximize the entropy of the output image to adjust the image brightness. This allows the module to primarily enhance low-light areas when the input enhanced image has low illumination, and primarily enhance high-light areas when the input enhanced image has high brightness. Compared to other image enhancement methods, this method has the advantage of increasing the brightness of low-light areas, increasing the contrast of extremely dark areas, decreasing the brightness of high-light areas, and increasing the contrast of over-illuminated areas in the processed image.

[0085] According to another aspect of the present invention, an image enhancement system for uncertain illumination is also provided, comprising,

[0086] The guided filter layering module is used to enhance the original image by dividing it into a base layer and a detail layer;

[0087] The detail layer enhancement adjustment module is used to stretch the detail layer based on the brightness of the base layer to obtain an enhanced detail layer; then the stretched and enhanced detail layer is blended with the base layer.

[0088] The adaptive brightness adjustment module uses an image entropy-based adaptive brightness adjustment method to adjust the fused image to obtain the final image with adjusted brightness.

[0089] The image to be enhanced, as input to the algorithm, first enters the guided filtering-based image layering module and is decomposed into a base layer I. baseand detail layer I detail Output base layer and detail layer I detail Send it to the detail layer enhancement adjustment module, and the module adjusts it according to the base layer I. base Brightness of detail layer I detail Arctangent stretching enhancement was performed to obtain the enhanced image I. detail_p Stretch and enhance the detail layer I detail_p With base layer I base The enhanced image I is obtained by fusing the components. enhance Enhanced image I enhance The image I is fed into the adaptive brightness adjustment module and subjected to an adaptive curve guided by "maximum entropy" to enhance it. enhance The brightness is adjusted to obtain the final output image I. out .

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention. The sequence numbers of the above embodiments of the present invention are merely for description and do not represent the superiority or inferiority of the embodiments.

Claims

1. An uncertain illumination image enhancement method, characterized in that, Comprise: S1: using a guided filter method to decompose the image to be enhanced into a base layer and a detail layer; S2: according to the brightness of the base layer, the detail layer is stretched to obtain an enhanced detail layer, and then the enhanced detail layer and the base layer are fused to obtain an enhanced image; S3: using an adaptive brightness adjustment method based on image entropy to adjust the enhanced image to obtain the final output image; specifically comprising the following steps, S301: output image after adaptive brightness adjustment The dispersion of brightness, i.e. image entropy, determines the gamma transformation parameter that enables the output image to have maximum entropy ; S302: According to the enhanced image The log average of the illumination intensity, determine the adaptive parameters used in the classification and transformation process ; By designing adaptive parameters to enhance the image When the overall illumination is low, mainly enhance the area with lower illumination, in the enhanced image When the overall illumination is high, mainly enhance the area with higher illumination, the design of adaptive parameters is: ； wherein, is an adaptive parameter, N is the number of pixels of the input image, is a very small positive number, is the enhanced image the gray level of the pixel; S303: dividing all pixels into high-luminance pixels and low-luminance pixels according to the adaptive parameter x' luminance, and all pixels are divided into high-luminance pixels and low-luminance pixels according to the adaptive parameter x' The pixel division is described as follows: ； wherein, is an adaptive parameter, is a gray level of an enhanced image pixel, is a transform function for different gray level pixels, is an output image pixel after pixel transformation; The transformed function after pixel division is as follows: ； wherein, is an adaptive parameter, is a gray level of the enhanced image pixel, is an enhanced image a result of the pixel transformation at different gray levels; S304: Perform the gamma transform on the two types of pixels according to the determined gamma transform parameters and adaptive parameters Perform the brightness transform on the two types of pixels in step S303 respectively to obtain the final output image .

2. The uncertain illuminance image enhancement method of claim 1, wherein, In step S1, the decomposition process is represented as: ； wherein, is the image to be enhanced, is the guided filter transform function, respectively denote the detail layer and the base layer obtained by guided filter decomposition.

3. The method of claim 1, wherein the step of enhancing the image is performed by using a non-deterministic illumination model. In step S2, the stretching enhancement function for stretching the detail layer image according to the base layer which can represent the background light intensity is: ； wherein, denotes a detail layer and a base layer obtained by guided filter decomposition, denotes a detail layer after stretch enhancement processing.

4. The method of claim 1, wherein the step of enhancing the image comprises the step of: In step S2, the fusion process is represented as: ​ ； wherein, represents a base layer obtained by guided filter decomposition, represents a detail layer after stretch enhancement processing, represents an enhanced image, represents a fusion coefficient of the detail layer after stretch enhancement processing in a fusion process. ​ 5. An uncertain illumination image enhancement system for implementing the method according to any one of claims 1-4, characterized in that, Comprise: Guided filter layering module: using a guided filter method to decompose the image to be enhanced into a base layer and a detail layer; Detail layer enhancement adjustment module: according to the brightness of the base layer, the detail layer is stretched to obtain an enhanced detail layer, and then the enhanced detail layer and the base layer are fused to obtain an enhanced image; Adaptive brightness adjustment module: using an adaptive brightness adjustment method based on image entropy to adjust the enhanced image to obtain the final output image.

Images