Research On Enhancement Methods Of Low-light Color Image

The detailed information of the image has important application value.The quality of the acquired image is often degraded due to weather or uneven illumination.Low contrast image cannot meet people's needs for high-definition images.Image enhancement algorithms can improve the visual effects of low-light images and enhance the details of the image.This thesis mainly focuses on the research of low-light color image enhancement methods.The specific research content is as follows:1.In this thesis,the theories of image enhancement algorithms based on histogram equalization(HE),Retinex algorithm,multi-scale transformation and deep learning are introduced.The advantages and limitations of these algorithms are compared and analyzed.The basic principles of image enhancement methods based on HE and Retinex are discussed.2.A color image enhancement algorithm based on weighted histogram equalization is proposed.The image pixel gray value is adjusted adaptively by introducing the weight parameters.A new adaptive mapping function is constructed.Performing two different mappings of image gray levels can effectively overcome gray level merging and loss of details.And the input image is stretched by color amplitude and saturation to complete the restoration of image color information.3.A multi-scale Retinex with color restoration factor(MSRCR)image enhancement algorithm based on adaptive weight is proposed.In the HSV color space,the brightness channel image is decomposed into Retinex enhancement layer and detail restoration layer based on weight.Retinex algorithm that uses the probability distribution characteristics of pixels to obtain adaptive weights to calculate the reflection component.The image in detail restoration layer is processed by guided filtering.The image in smoothing layer with basic information of the image will be obtained.The difference between the original image and the image in smoothing layer can be used to obtain a image in edge detail layer containing image contour information.The image in edge detail layer enhanced by the gain coefficient and the image in smooth layer and can be recombined.The enhanced image in detail recovery layer is aquired.The enhanced Retinex enhancement layer and the detail restoration layer are fused to output the final result image.The gamma correction algorithm can restore part of the lost detail and color information during the fusion procedure.The algorithm effectively eliminates the whitening phenomenon in the enhanced results of the traditional multi-scale Retinex algorithm.The simulation experiment results show that the algorithm proposed in this thesis has achieved satisfactory results in both subjective perception effects and objective evaluation indicators.The algorithm has practical application value.