Fuzzy Logic-Refined Color Channel Transfer Synergism based Image Dehazing
Sriparna Banerjee, Shambhab Chaki, Soham Jana, Sheli Sinha Chaudhuri · 2020 IEEE Region 10 Symposium (TENSYMP) · 2020
This paper introduces a novel Refined Color Channel Transfer (RCCT) prior as an improved alternative of existing Color Channel Transfer (CCT) prior. Like CCT, RCCT also compensates the chromatic losses occurring in degraded hazy images by employing a global color transfer strategy but it performs color transfer using well-scaled reference images generated using our proposed Fuzzy logic based reference image generation technique in contrary to CCT which usually performs color transfer using reference images possessing over-enhanced glow (bright) regions and poorly enhanced lowlight regions. The presence of such over-enhanced /poorly enhanced regions in the references images used by CCT significantly affect the visibility of outputs obtained from the dehazing methods where CCT acts as a pre-processing step. To overcome these shortcomings, here we have proposed a novel Fuzzy logic based reference image generation technique which restricts the intensities of generated reference images within allowable ranges by introducing a control parameter `k'. A unique value of `k' used for controlling the intensity of each pixel is computed depending upon the properties of the super-pixel in which it belongs, using a novel set of Fuzzy Inference (FI) rules which facilitates the production of visually improved outputs and also enables RCCT to serve as an ideal preprocessing step of various daytime, nighttime and underwater dehazing methods which is experimentally proven in this work.