SINGLE IMAGE FOG REMOVAL USING IMPROVED DARK CHANNEL PRIOR AND CONTRAST LIMITED ADAPTIVE HISTOGRAM EQUALIZATION

Abstract

One of the major problems in image processing is the visibility restoration of images corrupted by atmospheric degradation such as fog, rain and snow. In outdoor environment when images are captured from optical devices, light after reflecting from an object is scattered in the atmosphere before it reaches the camera. This is due to the presence of atmospheric impurities in the air in pollutants and atmospheric degradations such as fog, rain and snow. These atmospheric impurities result in two fundamental phenomena called ‘direct attenuation’ and ‘airlight’ which create a problem of color ambiguity. ‘Direct attenuation’ reduce the contrast by attenuating the reflected light of distant object reaching the camera and ‘airlight’ add the whiteness in the scene. As a result, images taken under such conditions, appeared hazy with poor contrast and they lose their visual vividness and color fidility. Although this effect may be desirable from an artistic standpoint. But for outdor-vision application used for object recognition, tracking, navigation and surveillance, one may need to restore an image throgh a process generally referred as fog removal. Accordingly, the task of this thesis is to present a simple but effective method to remove fog from a single digital image. Proposed method operates at a high speed than current methods and can minimize halo-artifact. This new method uses “Guided filter” for refining of transmission map obtained by “Median dark channel prior” and “Contrast-limited-adaptive-histogram-equalization” as a post-processing operator to improve the overall contrast of image.