Single Color Image Defogging Based On Dark Primary Color Prior And Variational Model

Fog is a common natural phenomenon.It will degrade atmospheric visibility and the image of the scene,then seriously affect the outdoor image acquisition and processing.It is of great practical significance to study how to obtain clear images in foggy weather,and to enhance the restoration of atmospheric degraded images and the enhancement of detail information.Since dark channel prior can restore images with different degree of fog and scene depth,and the variational models can maintain the features such as image edges and textures better,four haze removal algorithms are proposed in this paper on the basis of dark channel prior and the variation model.Combined with dark channel prior and MTV(multichannel total variation)model,we proposed an improved algorithm called H-MTV model.Then,using auxiliary variables and Dual Bregman iterative parameters to calculate the mode.Similarly,combined with dark channel prior and CTV(color total variation)model,we proposed an improved algorithm called H-CTV model.Then,using Split Bregman algorithm to calculate the model.The variational models based on local operators often have the problems of edge blur and texture blur,when they are used in color images dehazing.However,the variation models based on the non-local operators can maintain the image features better.Therefore,the traditional MTV and CTV models are extended to the NL-MTV(non-local multi-channel total variation)and NL-CTV(non-local color total variation)models based on nonlocal operators for color image dehazing.Then,combined with the dark channel prior,H-NL-CTV model and H-NL-CTV model based on nonlocal operators are proposed.Finally,mathematical statistic methods are used to compare H-MTV model,H-CTV model,H-NL-MTV model and H-NL-CTV model,with subjective quality evaluation and objective quality evaluation.Furthermore,comparing the four proposed models with He algorithm and Kimmel Retinex algorithm,experimental results show that the proposed algorithm are superior to the traditional methods visually and objectively.