Research On Defogging In Single Image

The images captured in the foggy weather conditions will be severely degraded.The specific performance of the degradation can be described as declining in imagecontrast, blurring outline of scenery, diming color of the images and so on. This leadsto the fog images are not accord with human visual characteristic, and also notconducive to feature extraction by visual machine. Then the target detection andtracking, and other related work that based on image feature will not on the rails. Thus,it is very important to make the fog images distinct. The main research contents andcontributions of this paper can be summarized as follows:First of all, the image is defogged from the perspective of image enhancement,including histogram equalization, partially overlapped sub-block histogram andSub-block non-overlapping histogram equalization. This kind of method is thefundamental image enhancement method, and the result is often used to compare withother algorithms. Then defogged the image by algorithm that based on homomorphicfilter, image illuminance reflection model is established, after that, combinespectral-filter with gray transformation to enhance the image contrast and compressthe brightness range. Finally, methods based on single-scale Retinex and multi-scaleRetinex are analysed and realized, and the relevant algorithms are evaluated accordingto the result of the experiments.The other technique is creating a atmospheric scattering model from theperspective of the specific and physical causes of the images degradation, and thenestimate the atmospheric medium factor transmittance and light intensity on the basisof dark channel prior estimation. In other words, it models the degradation process.This method is usually referred to image restoration. Due to the rough transmittancelead to vignetting effect in the image, soft matting technology is used to transmittanceoptimization.These are traditional methods of defogging in single image, manyimprovements have been made to the above methods. But some improvementmethods improve the image processing speed while sacrificing the quality of theimage, and others improve the quality of the image at the expense of the processingspeed. This article is the first to apply the image enhancement method that based onpulse coupled neural network (PCNN) to defogging area. Firstly, the principle of gray image enhancement by PCNN is analyzed. Then combine it with the theory of colortransformation to enhance the luminance component in the HIS color space, thenimage color saturation is adjusted nonlinearly, finally the process end up with a clearimage. At the end of this article, the effectiveness of the algorithm has been proved bycomparison with traditional defogging algorithm in two aspects, subjective evaluationand the time used to defog.