Haze image restoration based on physical optics model using raspberry pi B+V1.2

Abstract

Dehazing is still a challenging problem for the outdoor images. The occurrence of haze, fog, mist etc., reduce the visibility of captured images in the outdoor environment. There are several dehazing techniques proposed to resolve the haze in the outdoor images but the existing techniques suffers from the computational time complexity. For a real-time application time complexity plays a major role. As a part of real-time application haze removal is done using the raspberry pi b+v1.2. The hazy image is restored by taking dark channel with patch size of 11 × 11 to avoid the oversaturation of recovered scene radiance and halo effect. Getting the image size and channels, the image is flattened to get the top bright points. Then the image is padded at the ends with a square of side of the window to get the minimum comparision at the ends. Because the atmospheric light to be taken from the foreground as the background has almost same pixel values distributed. The intensity is found, from that maximum among these pixels are found and considered for atmospheric light. The transmission map is found, hazy image is restored. And the performance of modified approach is compared with paper perception oriented haze image restoration based on physical optics model. Hence, to give a better dehazing images with reduced computational time using raspberry pi b+v1.2.