Improved Dark Channel Prior for Image Defogging

Abstract

An image defogging algorithm with improved transmittance is proposed for the problem of residual fog in the defogged image obtained by the dark channel a prior algorithm. In this paper, the atmospheric light value in the scene is estimated using the quadrature method, and the minimum map in the dark channel algorithm is further optimized using the super pixel segmentation algorithm with median filtering and combined at the pixel level. The obtained transmittance map is then filtered with a guide to eliminate texture effects. Finally, the defogged image is transferred to the HSI (Hue-Saturation-Intensity) color space for image enhancement. The results of the experiment showed that compared with the classical dark channel algorithm, the algorithm in this paper has obvious defogging effect, complete image information retention, and lower algorithm complexity.