Video Enhancement with Single Frame

Vijay M. Galshetwar, Prashant W. Patil, Sachin Chaudhary

Abstract: Hazy scenes captured by the camera in a hazy environment are the main cause of the deterioration of captured video quality. Hazy environmental conditions degrade the quality of captured videos which leads to poor visibility and color distortion in videos. In this paper, we propose a video enhancement technique with a single frame dehazing encoder-decoder network. The proposed method focuses to restore the haze-free video from the hazy video by developing an encoder-decoder-based dehazing network. The framework comprises of encoder block (EB) and decoder block (DB). Both EB and DB are used to extract the haze-relevant feature maps and to give more attention to haze-specific features by re-calibrating the weights of learned feature maps automatically and helps to recover the haze-free frame. The use of dilated residual module (DRM) helps to improve output results and gain more contextual information by widening the receptive field. The extensive analysis of the proposed method is carried out by considering benchmark synthetic hazy video databases and analyzed quantitative results. Experimental results show that the proposed method out performs the other state-of-the-art (SOTA) existing approaches for video frame dehazing.