Sand-dust image enhancement based on light attenuation and transmission compensation

The study of sand-dust images has made remarkable progress in recent years. However, it is challenging to balance the image quality, color casts correction, and running time well in the existing sand-dust image processing methods. This paper introduces a novel compensation coefficient and effective intensity difference prior, which leads to an efficient and robust sand dust removal method. This method assumes that the medium transmission coefficients of each pixel are not equal. First, the rough transmission is estimated by the red-green channel pixel-wise, thereby leading to significantly reduced running time. Then, the difference between the blue channel and red-green channel is utilized to compensate for the rough transmission. Besides, ambient light is estimated using the minimum absolute intensity difference between channels, according to the attenuation characteristics of light in sand dust. Meanwhile, a color adjustment method based on global stretch and the green channel is also improved, making the method effective on various sand-dust images. A series of experiments are conducted on a number of challenging sand-dust images with the proposed method and other state-of-the-art sand dust removal techniques, revealing the superiority of the proposed method in terms of calculation time, color shift correction, and restoration quality over all the comparable techniques.