A low illumination video enhancement algorithm based on the atmospheric physical model

In this paper, we propose a fast and effective low illumination video enhancement algorithm by combining the Retinex theory with the dark channel prior theory to improve contrast and reduce noise of low illumination videos. Considering enhancing low illumination videos and amplifying noise simultaneously, noise reduction before enhancement is beneficial to improving video enhancement effects. Therefore, we combine advantages of guided filtering and median filtering to propose an improved comprehensive de-noising algorithm, which would be applied to YCbCr space. Then, we quickly estimate luminance transmission maps in HSI space and apply the atmospheric model to recover the low illumination video. Finally, the scene detection and the inter-frame compensation are utilized to further improve effectiveness and speed of the process. Experimental results show that the proposed algorithm could effectively improve luminance and contrast of low illumination videos, reduce noise and strengthen detailed information of videos and make sure the continuity of the inter-frame motion, thereby improving the quality of videos.