PROJECT TITLE :

A Hierarchical Approach For Rain Or Snow Removing In A Single Color Image - 2017

ABSTRACT:

During this paper, we have a tendency to propose an economical algorithm to remove rain or snow from one color image. Our algorithm takes advantage of 2 popular techniques utilized in image processing, namely, image decomposition and dictionary learning. At first, a mix of rain/snow detection and a guided filter is employed to decompose the input image into a complementary pair: one) the low-frequency part that is free of rain or snow nearly completely and 2) the high-frequency half that contains not only the rain/snow element but also some or perhaps several details of the image. Then, we tend to focus on the extraction of image's details from the high-frequency half. To this finish, we tend to style a three-layer hierarchical scheme. In the first layer, an overcomplete dictionary is trained and three classifications are administrated to classify the high-frequency half into rain/snow and non-rain/snow elements in that some common characteristics of rain/snow have been utilised. Within the second layer, another combination of rain/snow detection and guided filtering is performed on the rain/snow component obtained in the primary layer. Within the third layer, the sensitivity of variance across color channels is computed to reinforce the visual quality of rain/snow-removed image. The effectiveness of our algorithm is verified through both subjective (the visual quality) and objective (through rendering rain/snow on some ground-truth images) approaches, that shows a superiority over several state-of-the-art works.