Research Of Enhancement Algorithm For Low-light Image Based On Dark Channel Prior

Under the condition of insufficient light source,the captured images usually have a lower recognition degree,which is called low illumination image.Low-light images are common,usually with problems like color distortion,low contrast and so on,which make it difficult for the naked eye to distinguish the content of the image.Genera lly,algorithms and applications of computer vision are based on high quality images.Low illumination images bring huge challenges to image analysis and processing,which seriously affect the efficiency of algorithm.There are two main methods of traditional image enhancement: airspace method and transform domain method.The typical spatial domain method has histogram equalization algorithm and Retinx series algorithm,and the transform domain method usually uses wavelet transform and homomorphic filtering.Although many image enhancement technologies have been put forward to solve the problem of low illumination,the above methods will reduce or increase the contrast more or less.In addition,because of the lack of physical modeling,the traditional methods are not very versatile,and it is easy to produce the problem of detail loss and color distortion.In order to solve this problem,Dong proposes a simple and efficient low illumination algorithm based on fogging algorithm.The algorithm uses the similarity between the low illumination reversal image and the fog image,and uses the dark channel prior technology to enhance the low illumination image.Although the algorithm can better enhance image contrast,block effect introduces noise,reduces image detail,and produces severe distortion when the dark channel is priori failure.According to the similarity between low illumination image inversion and foggy map,combined with the dark channel prior algorithm,we improved the algorithm of Dong.We use luminance information to calculate transmissivity map,and guided filter to optimize the result.Finally,information entropy,PSNR and SSIM are used as quality evaluation criteria to verify the improved algorithm.results of experiments prove that the method is effective and efficient,and the optimized transmittance map reduces noise and achieves more accurate brightness and detail enhancement.