WITHDRAWN: Application of CNN and gated recurrent network for visual improvement of dehazing

Abstract

Abstract Various restoration model depending on the prior knowledge had consider an arduous vision task. We present in this paper an effective algorithm for restoring a recognizable picture from a hazy image. Instead of using conventional methods i.e.; image priors as the restoration process, in form of color attenuation, dark channels, or color line contrast, we use a point-to-point gated background aggregation structure to reestablish the final image free from haze. A hybrid dehazing algorithm of CNN framework with features learning recurrent modeling is applied. The transmission map or its depth of scene, structural attributes is obtained in form of ranking layer is applied with the CNN network model so that hazy images can be captured at same instant. Post processing is carried out by Fast guided filter. Hence this hybrid model basically gives a inter connection idea between coarse layer and fine layer by extensive results obtained on experimental, simulated and real-world images prove that the proposed algorithm operates against state-of-the-art algorithms in a beneficial way.