A Multi-Scale Decomposition Based Haze Removal Algorithm

The contrast and colors of images acquired in bad weather are usually poor. This paper proposed a simple method to remove haze based on a multi-scale decomposition method. Through the Fourier spectrum and Hough transform we obtain the directional images which are then used to decompose the image to IMFs. By a linear mapping, we manipulated the detail, luminance as well as color to enhance the degraded images. Experiments show the efficiency of the proposed method.