Content-Based Image Retrieval System
Overview
For this project we created a system for content-based image retrieval. The system employs a query-refine strategy, where images retrieved via a text-based query are refined based on shape similarity given user input. The shape similarity measure is based on representative shape contexts. We developed an efficient Hough transform-based method that preserves spatial coherence when comparing images. We conducted quantitative analysis and compared the method's effectiveness on Corel image and Google image datasets.
Collaborators
I worked on this project with professor Greg Mori and Chris Lundgren. I developed the web frontend that pulled images from Google image search, applied the RSC calculation, and presented the results and Chris developed the RSC calculations.