关键词: 低光照图像, 图像增强, 等分直方图均衡, 小波变换, 高频强化

Abstract: In low light situation, the imaging often suffers dark brightness and low contrast, the image information is not clear, and the existing image enhancement methods can not fully improve the brightness, contrast and clarity of the image. This paper proposes a low light image enhancement method based on equidistant histogram equalization and high frequency highlighting. Firstly, by virtue of the independence of brightness component and color component in HSI color space, the image is converted to HSI color space, and the brightness image is decomposed into low-frequency image and high-frequency image according to Retinex theory and Gaussian filtering. And then, an equal histogram equalization method is proposed for low-frequency image, which divides the frequency of intensity and the dynamic space of low-frequency image into four equal parts, respectively, and piecewise remaps the pixels of low-frequency image using the pixel frequency of local intensity level, and Retinex reconstruction with low-frequency and high-frequency images is performed, so as to improve the brightness and contrast of images. Finally, the enhanced brightness image is transformed to the wavelet domain, and the components of high-frequency wavelet coefficients are increased in a fractional way to improve the clarity of the image and highlight the edges and details of the image. Experiments on low light image datasets show that the images enhanced by the proposed method in this paper are better than some existing methods in enhanced image, average gradient, information entropy and edge extraction. The method in this paper can increase the application value of image more effectively.

Key words: low light image, image enhancement, equidistant histogram equalization, wavelet transform, high frequency highlighting