Abstract: In order to enhance the image contrast and compensate the loss of color information simultane- ously, this paper proposes a novel single image dehazing method based on entropy and information fidelity. Firstly, the global atmospheric light is estimated by quad-tree form transformed hazy image. Then, transmis- sion is estimated by a joint objective function based on entropy and fidelity at non-overlapped sub-block re- gions. It is further refined by weighted least squares optimization to alleviate block artifacts. Finally, dehazed image is constructed from atmospheric scattering model. Real haze outdoor images under different scenes and resolutions are used as test data. Moreover, multiple evaluation indicators are employed to syn- thetically analyze the performance of proposed method. The comparison experiment results show that pro- posed approach performances better than other conventional methods in many evaluating indicators espe- cially in colorfulness and global contrast factor.