In order to improve the level of video image clarity in non-uniform low-light environment, a research method of video image clarity in non-uniform low-light environment is proposed. A video dynamic image illumination change scale decomposition model is construct, and according to that sensitivity of the illumination initial transformation parameter, the noise feature parameters of the matched video dynamic images are fuse and the outlier parameters are decomposed. A spatial uniform noise filtering model is established by using a high-order similarity feature matching method, and a video dynamic image is made clear according to a filtering result. The experimental results show that the signal to noise ratio (SNR) of the image processed by this method reaches 47.8 dB, the mean value of image structure similarity is 0.980, the mean square error is only 0.086, and the processing time of image clarity is 5.21 s, which proves that this method is practical and can effectively improve the ability of image clarity.