Removing Atmospheric Noise Using Channel Selective Processing For Visual Correction

Abstract

In the presented paper; we propose an effective image fog removal technique with a color stabilization technique which is a total 2-level process for image restoration with a HSI (Hue Saturation Intensity) based evaluation process. The approach uses extraction of suppressed pixels from an RGB image affected by smoke, steam, fog which is form of white and Gaussian noise. From our observation of most images in fog environment contain some pixels which have low values of luminescence in every color channel (considering RGB image).Using this model, we can directly estimate the effective density of fog and recover the most affected parts in the image. The parameter of calculating the effective luminescence which is a form of intensity, and also gives the scattering estimates of the light, the combined Laplace of the luminescence-light and suppressed pixels values gives us the basic map of light spread which is further used in the restoration of intensity. The transmission of intensity between the calculated fog values in the image give the estimate for the local transition between the intensity values and color values. This factor helps in the color restoration of the affected image and estimates the proper restoration of image after removal of dense fog particles. After the removal of fog particles, we then restore the color balance in the image using an auto-color-contrast stabilization technique. This is the 2-level fog restoration method. The visibility is highly dependent on the saturation of color values and not over saturation, which accounts for image quality improvements. In order to evaluate in-depth the effectiveness, we have also introduced the HSI mapping of the images, as this will show the true restoration of intensity and saturation in the fog image. Results on various images demonstrate the power of the proposed algorithm. To measure the efficiency of the algorithm the parameter of visual index is also estimated which further evaluates the robustness of the proposed algorithm for the HVS (Human Visual System) for the de-fogged images.