Image Enhancement Research Based On Retinex Threshold Segmentation Algorithm

The image enhancement algorithm based on Retinex theory has better image enhancement effects for uneven illumination and low contrast images,and has a wide range of application in various industries.The traditional Retinex image enhancement algorithm is improved in the process of image enhancement,such as edge information loss,halo,graying,and so on.This paper proposes a Retinex algorithm based on threshold segmentation.Firstly,with the help of the guide filtering enhanced by Canny operator,the light image is estimated,and the edge information is strengthened.Secondly,the multi-threshold interval OTSU threshold segmentation algorithm is used to segment the reflected image into bright and dark regions,which improves the accuracy of segmenting the bright and dark regions.Using image information entropy contrast and gradient ascent method,the quantization parameters and Retinex illumination estimation parameters of the reflected image are adjusted to select the optimal image corresponding to the bright and dark area.Finally,the DCT algorithm and different fusion strategies are used to fuse the optimal images in the light and dark regions,which can effectively eliminate the phenomenon of graying in the region and achieve the optimal image enhancement effect.In this paper,the algorithm is verified by haze images,low-contrast images and uneven illumination images.The results of comparative experiments show that the information entropy,average gradient and standard deviation have been improved,and the algorithm in this paper is effective and has better enhancement effect.The paper has 33 figures,4 tables and 57 references.