Image Segmentation and Adaptive Contrast Enhancement for Haze Removal
Chunyan Wang, Bao Liang Zhu · 2020
With a view to restoring image details of heavily hazy images, we propose an adaptive contrast enhancement algorithm specifically for haze removal. It is composed of 3 parts. The first part is to segment the input image into flat background of air space and foreground which is the rest of the image. A specific gradient matrix is defined to generate a gradient feature value to identify the pixels of very weak signals with the presence of noise of similar amplitude. In the second part, a CLAHE-based method is developed and applied to the foreground to provide a stronger enhancement to weaker signal variations while the background is protected from noise enhancement. A specifically designed filter is then applied to remove noise caused by the discontinuity between the foreground and background areas, while preserving the enhanced image details. The proposed algorithm has been tested and its effectiveness has been proven by the test results.