Research On Natural Image Dehazing Algorithm And Clarification Objective Evaluation

In recent years,with the continuous development of computer technology,image processing technology has been widely used in various fields including industry,aerospace,military and life.However,due to the presence of random media such as fog and haze in the atmosphere,resulting in blurred image details,low contrast,dim colors when shoot images in a foggy environment,then greatly reducing the application value of the image.Therefore,it is of great significance to dehaze the foggy image to make the image clearer.This dissertation first introduces the relevant theory of the defogging algorithm,and then based on this,researches a single image defogging algorithm based on improved guided filtering.Finally,from the perspective of human visual perception,this paper studies a clearness measurement method for defogged and blurred images by using a non-reference quality evaluation method.The specific research is as follows:(1)In order to improve the defogging effect of foggy images,aiming at the problems of unnatural colors and inconspicuous outlines of objects in the defogged image after optimizing the transmission rate by using a guided filter,a new image dehazing algorithm based on improved guided filtering is obtained.This paper firstly proposes an optimization method for averaging atmospheric light values based on pixel mean.Then,by introducing a first-order edge perception factor and a pixel position perception factor,a more accurate transmittance is obtained after improved guided filtering.The experimental results show that the algorithm can effectively correct the defogging effect of images containing white areas,and can also effectively improve the phenomenon of incomplete defogging,and the contour details of the image after defogging are significantly enhanced.(2)Combining visual features with underlying features,a method of extracting salient regions of an image by combining SDSP model and adaptive threshold binarization is studied.Experiments on ASD and ECSSD datasets show that this method works well.(3)Because the traditional blur estimation algorithm focuses on the edge area,the complexity and difference of the image content are not fully considered,and the importance of human visual attention is also ignored,resulting in lower objective score than subjective score.Aiming at the above problems,a saliency weighted and fused image sharpness model is proposed.From the perspective of human visual perception,the model incorporates the local detail variance of global sharpness and re-fuzzy theory based on CPBD,and adds a significant weighting factor to reflect the significance of important areas.The experimental results show that the objective results of the image sharpness function proposed in this paper are more consistent with human subjective feelings.