Research And Implementation Of Low Illumination Image Enhancement Algorithm Based On Retinex Theory

As a carrier of information,images have been widely used in the fields of communications and computers.Among the massive images,low illumination images account for a large proportion due to the shooting environment.Low illumination image generally contains obvious noise,lower brightness,less details and severe color distortion.The above disadvantages reduce the application value of low illumination images.Low illumination image enhancement processing is an important branch of image enhancement.Traditional enhancement algorithms can only improve some of the disadvantages of low illumination images,and the enhancement effect is not ideal.Retinex algorithm has color constancy,and can simultaneously achieve brightness enhancement,detail enhancement and color fidelity,which improves the quality of low illumination image more comprehensive.In this paper,a multi-scale Retinex algorithm with adaptive weights is proposed for images with low overall brightness.Low illumination images contain more zero pixels,so amplitude compensation mechanism is introduced to solve the problem of truncation of image data caused by logarithmic transformation in Retinex algorithm.The image is divided into blocks,and the best scale parameters of the sub block are calculated according to the flat characteristic of the image.The adaptive weights are calculated using the relationship between the optimal scale parameter and the classical scale parameters,which ensures the best enhancement effect in the sub-block,and solves the problem of poor performance of multi-scale Retinex algorithm using the global filter.The algorithm uses the relationship between the color channels to define the normalized brightness of the image,and then uses the normalized brightness to linearly match the low illuminance original image and the enhanced result,avoiding the problem of over-enhancement of the normal brightness area of the image.Experimental results show that the algorithm in this paper works well in detail enhancement,brightness maintenance and color fidelity.A low illumination image enhancement algorithm based on PLIP(Parameterized Logarithmic Image Processing)model is proposed for low illuminance images containing brighter regions.The algorithm combines the operation rules of the PLIP model and the image convolution operation to form a Gaussian filter under the PLIP model.The algorithm processes the image brightness component in the HSV space,effectively reducing the value interval problem in the calculation of the reflection component.The algorithm makes use of the relationship between the background brightness of the image and the human visual system to make non-linear adjustments to the enhanced luminance component,which improves the visual effect of the image.According to the correlation between saturation and brightness,the algorithm corrects the saturation and improves the color perception of the image.Experimental results show that the algorithm in this chapter corrects the saturation of the highlighted areas while enhancing the low illumination area,making the enhanced results of the low illumination image more consistent with the human visual effect and avoiding over enhancement phenomenon.