Investigation of the Relative Radiometric Normalization in Remote Sensing Image Change Detection

In the change detection of remote sensing images, the brightness and colors need to be adjusted at first in order to make two images almost have the same radiation (relative radiometric normalization (RRN)). However, hazes in images often influence the effect of relative radiometric correction. In this paper, we proposed a new approach to detect and remove hazes from high-resolution panchromatic images based on undecimated discrete wavelet transform before RRN. The procedure consists of four steps: construction of haze thickness mask, processing in the high frequency and histogram matching. It can detect the haze areas and estimate their thickness form the scene by analyzing the low frequency information content, and rectifying the high frequency content by different amplitude according to the haze optical thickness. This method was tested in Tianjin, China, using Quickbird satellite data. Experimental results demonstrate that the proposed method is suitable for haze in image.