Removing atmosphric noise using channel selective processing for visual correction
Rajbeer Kaur · 2014
In this paper, we propose an effective image fog removal technique from a single input image. The approach uses extraction of minimized values of statistics of the fog-free outdoor images. It is based on a key observation–most images in fog-free outdoor images contain some pixels which have low values of luminescence in at least one color channel. Using this model, we can directly estimate the effective density of fog and recover a high quality fogfree image. The parameter of calculating the effective light intensity also gives the scattering estimates of the atmospheric light, the combined Laplace of the air-light is and minimum 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, this factor helps in the color restoration of the affected image and estimates the proper restoration of image after removal of dense fog particles. The visibility is highly dependent on the saturation of color values and not over saturation, which accounts for image quality improvement. Results on various images demonstrate the power of the proposed algorithm.