Feb 2023

As a class project, the objective was to gain a deeper understanding of pattern recognition, and subsequently content-based image retrieval. I achieved this by implementing baseline matching, RGB, rg chromaticity histograms, multi-histograms, and Sobel magnitude texture,  and experimenting with a few other methods and their combinations like Laws texture (L5E5), and Gabor filters for texture extraction. I tried to compare different distance metrics viz. sum of squared difference, histogram intersection, Chi-square, and Correlation distances. A set of training images were also chosen to test the various mix and match of feature calculation and matching metrics.

Please find the report for the same below.

Github: https://github.com/devvaibhav455/opencv_practice/tree/main/Project_02_Content_Based_Image_Retrieval