Visibility Enhancement of Fog Degraded Image Sequences on SAMEER TU Dataset Using Dark Channel Strategy
Tannistha Pal · 2018
Image sequences captured in poor visibility condition especially in foggy weather severely affect the scene structure of an image, which in turn will affect several computer vision applications like remote sensing, intelligent vehicles, and visual surveillance detection, tracking, and recognition of targets. Thus, retrieving back the clear scene from such foggy image sequences is of great importance. The main focus of this paper is to summarize current image dehazing algorithms and also implement a robust technique for increasing the visibility of the fog degraded image sequences. In this paper, we investigated different dehazing methods and classified them. After investigation of different dehazing approaches, we deeply analyze Dark Channel Prior Technique(DCP) and implemented this technique on Benchmark and on our own developed Dataset called SAMEER TU(Society of Applied Microwave Electronics Engineering & Research-Tripura University) Dataset on images and videos. Experimental results demonstrate that this algorithm is efficient which can restore the contrast and color of the scene effectively, and thereby improves the visibility of the image. Finally, to identify the robustness of the technique used, we used different objective image quality assessment methods. Finally, we propound the problems of defogging approaches which need to be further studied.