Image Fog Removal And Its Application Research Based On Soc

Real environment often encounter types of complex weather.Leading to the degradation of image,the efficient defog algorithm for the computer vision system is unreliable while recognition and understanding.Our works often deal with vehicle road images for active safe system,So the corresponding improvement and implementation for defog algorithm based on SOC includes as follows:1.A new defog algorithm based on L0 Smoothing guide filtering has been proposed by this paper.For vehicle road images with HDR perspective,First our obtain a row transmission image under multi-scale dark images as a input image for guide filter.out calculate a fine transmission image under a point scale dark channel image,Then using L0 Smoothing filter deal with it as a guided image for guide filter.To make sure that the final transmission image can keep the main edge information while guaranteeing depth information of the image accurately,guided filter use local linear model to fuse low frequency and high frequency information effectively between a input image and a guided image.2.Accelerating the defog algorithm,a new method uses high-level synthesis tool named HLS to design IP core has been proposed.Then the SOC platform for defog algorithm will be build;According defog algorithm based on classic dark prior channel and the excellent characteristics of guide filter,as well as the different computing architecture system and characteristics of SOC based ARM and FPGA.Finally our take HLS tool to implement hardware IP core accelerating defog algorithm Using SIMD computing architecture as follows: minimum filter IP,guided filter IP.3.A new defog evaluation of image quality based on SVM has been proposed by this paper.Solving the model parameters of a SVM linear classifier,our use the color and clarity characteristic of image make up 9 dimensional feature vector.Then our obtain the optimal hyperplane segmentation for SVM.As a comprehensive evaluation for the quality of image.Our experimental results show the good performance.