CN-121981929-A - Image enhancement method applied to complex scene

CN121981929ACN 121981929 ACN121981929 ACN 121981929ACN-121981929-A

Abstract

The invention discloses an image enhancement method applied to a complex scene, which is used for comprehensively treating the problems of color deviation, brightness attenuation and insufficient contrast of an input image through the steps of color space conversion, color channel modulation, brightness compensation, global contrast optimization, local brightness balance optimization and the like. The method comprises the steps of firstly converting an input image into a preset color space, carrying out nonlinear adjustment on hue and saturation components, then judging a spectrum attenuation direction through the average brightness relation of a color channel, carrying out brightness compensation, and realizing brightness smoothing through a neighborhood statistical mode. The method further performs contrast optimization based on global luminance statistics and promotes local region details using luminance histogram redistribution and local luminance factors. And finally, fusing the global enhancement result and the local enhancement result to obtain an enhanced image with natural color, coordinated brightness and clear details. The method is suitable for image enhancement processing under underwater imaging, low-illumination imaging and multi-type degradation scenes.

Inventors

CHENG JIEREN
WANG CHENGCHAO
TANG XIANGYAN
ZHANG ZIHUI

Assignees

海南大学

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (10)

1. An image enhancement method applied to a complex scene, characterized in that the method performs adaptive color correction and multi-granularity contrast optimization on an input image according to a preset process flow based on color properties, brightness properties and contrast distribution of the input image to generate an enhanced image, wherein the method comprises the following steps: Performing color space conversion on an input image to obtain a tone component, a saturation component and a brightness component, and performing color channel modulation based on the relation between tone and saturation to enhance a weak color region; calculating average brightness of each color channel of the input image, judging a spectrum attenuation direction based on brightness difference, performing brightness compensation on the color channel with brightness lower than that of the reference channel, and performing neighborhood smoothing on a compensation result to enable the brightness of the color channels to be consistent; step three, global contrast optimization is executed based on global statistical characteristics of the brightness channels, and the global entropy constraint and the brightness adjustment factor are constructed, so that the whole brightness range of the image is expanded, and the global layering sense is improved; Step four, local brightness balance optimization is executed based on local statistics of brightness channels, local contrast is adjusted by utilizing brightness histogram redistribution and local brightness factors, dark part details are enhanced, and bright part areas are kept balanced; And fifthly, carrying out weight fusion on the global enhancement result in the step three and the local enhancement result in the step four to obtain a final enhancement image.
2. The method according to claim 1, wherein the step one of performing color space conversion on the input image is to convert RGB color space of the input image into HSV color space, the hue component is used to represent a color type, the saturation component is used to reflect color purity, and the brightness component is used to describe image shading information.
3. The method of claim 1, wherein the color channel modulation in step one includes performing a non-linear mapping of hue components to map hue shift regions to hue values closer to a reference direction and performing an enhancement adjustment of saturation components to increase the color intensity of weak saturation regions.
4. The method of claim 1, wherein the brightness compensation in step two comprises: And respectively calculating average brightness of the red channel, the green channel and the blue channel, judging the channel to be compensated according to the average brightness, and carrying out brightness improvement on the channel to be compensated based on compensation factors, wherein the compensation factors are constructed according to preset brightness ratio.
5. The method of claim 1 wherein the smoothing in step two uses a median based on spatial neighborhood statistics to maintain the brightness variation continuous over the spatial distribution by ordering the brightness distribution of the compensated pixels in their neighborhood and taking the median.
6. The method of claim 1, wherein the global contrast optimization in step three comprises computing a global luminance probability distribution for luminance channels and constructing a global contrast optimization factor based on luminance entropy, expanding a luminance dynamic range and stabilizing a luminance variation by performing linear expansion and luminance constraint on the luminance channels.
7. The method of claim 1, wherein the optimizing of the local luminance balance in step four comprises: And performing histogram equalization based on the cumulative distribution function on the brightness channel, taking the ratio of the equalized brightness to the original brightness as a local brightness factor, and adjusting the brightness and the contrast of the local area according to the local brightness factor.
8. The method of claim 1, wherein the fusing in step five is: and (3) respectively calculating a global brightness factor and a local brightness factor for the enhancement results of the third step and the fourth step, and performing linear fusion on the global brightness factor and the local brightness factor according to a preset weight coefficient to obtain an enhanced image with natural brightness and clear details.
9. The method of claim 1, wherein the method performs normalization processing on the input image prior to processing the input image to maintain the color channel pixel values in a uniform interval to ensure stable parameter ranges during subsequent color modulation, brightness compensation, and contrast optimization.
10. The method of claim 1, wherein the method includes underwater imaging, low-light imaging, or imaging affected by haze for input images in different complex environments, and performs color correction and contrast enhancement processing on the images without changing parameter settings.

Description

Image enhancement method applied to complex scene Technical Field The invention relates to the technical field of image processing, in particular to an image enhancement method applied to a complex scene. Background The image is easily affected by various factors such as color shift, brightness attenuation, insufficient contrast, detail weakening and the like in a complex imaging environment, so that the imaging quality is obviously reduced. In underwater, low-illumination and haze-affected environments, images often exhibit significant color cast due to differences in spectral absorption and scattering effects during light propagation, with the attendant problems of overall brightness imbalance and local area detail blurring. To enhance the visual quality of such images, image enhancement techniques typically require a multi-dimensional integration of color correction, luminance recovery, and structural detail enhancement to improve the visual representation and subsequent analysis performance of the image. The existing image enhancement methods mainly comprise an image restoration model-based method, a histogram or brightness distribution adjustment-based method and a deep learning-based end-to-end enhancement method. The recovery model method generally relies on inversion imaging degradation process, but is sensitive to imaging conditions, and is easily influenced by factors such as water quality change, light source difference and the like, so that an enhancement result is unstable. Although the conventional histogram equalization or brightness redistribution methods are simple to implement, the conventional histogram equalization or brightness redistribution methods often generate over-enhancement phenomena in a high-light area, and are difficult to effectively improve in a low-color or low-contrast area. The deep learning-based method can learn the enhancement strategy from a large number of samples, but is often limited by the diversity of training data and the generalization capability of a model in practical application, and artifacts or color shifts are easy to generate when facing complex scenes with color cast, low brightness and multiple degradation factors superimposed. Furthermore, existing approaches often have difficulty in simultaneously compromising color naturalness, brightness balance, and detail structural integrity, resulting in insufficient stability of the enhanced results when applied across scenes. In the prior art, when images in complex scenes are processed, the problems of unstable enhancement results, inconsistent brightness distribution and insufficient color correction still exist, and especially in imaging environments with superposition of different degradation types, the collaborative processing of color shift, brightness attenuation and local detail attenuation is difficult to realize. Therefore, there is a need for an image enhancement method for complex imaging conditions, which can stably improve color naturalness, brightness balance and structural detail performance of an image under different environments, and improve the overall visual quality of a complex scene image. Disclosure of Invention In order to solve the problems in the prior art, the embodiment of the invention provides an image enhancement method applied to a complex scene. The technical scheme is as follows: In one aspect, there is provided an image enhancement method applied to a complex scene, the method performing adaptive color correction and multi-granularity contrast optimization on an input image according to a preset process flow based on color properties, brightness properties and contrast distribution of the input image to generate an enhanced image, wherein the method comprises the steps of: Performing color space conversion on an input image to obtain a tone component, a saturation component and a brightness component, and performing color channel modulation based on the relation between tone and saturation to enhance a weak color region; calculating average brightness of each color channel of the input image, judging a spectrum attenuation direction based on brightness difference, performing brightness compensation on the color channel with brightness lower than that of the reference channel, and performing neighborhood smoothing on a compensation result to enable the brightness of the color channels to be consistent; step three, global contrast optimization is executed based on global statistical characteristics of the brightness channels, and the global entropy constraint and the brightness adjustment factor are constructed, so that the whole brightness range of the image is expanded, and the global layering sense is improved; Step four, local brightness balance optimization is executed based on local statistics of brightness channels, local contrast is adjusted by utilizing brightness histogram redistribution and local brightness factors, dark part details are enhanced, and bright part