Tensor-based weighted least square decomposition haze removal algorithm
Xin Jin, Xiaotong Wang, Xiaogang Xu, Chengtao Yi, Changqing Yang · 2017
Haze removal is a challenging work in outdoor image applications. Physical model based restoration methods are accepted with higher pertinence, while non-physical model based enhancement methods are more robust and widely applied. A novel haze removal algorithm based on tensor weighted least square decomposition was presented in this paper. By either progressively or recursively applying this decomposition, a set of multiscale outputs and differences were obtained. Then haze images were got dehazed by suppressing the haze layer while enhancing the extracted detail layers. The effectiveness and robustness of our haze removal algorithm were demonstrated by comparing our results with existing generally acknowledged dark channel prior based method.