Implementation of Content Based Image Retrieval Techniques for Video Recognition Applications

Document Type

Project

Lead Author Type

CIS Masters Student

Advisors

Dr. Jonathan Leidig, jonathan.leidig@gvsu.edu

Embargo Period

5-19-2016

Abstract

The goals of this project were to 1) identify a multi-media analysis pipeline, 2) research various methods of CBIR that could be applied to this problem, 3) identify the challenges inherent in the process of capture, process, and comparison, and 4) become familiar with image processing as a whole. This was done by simulating distortion of captures and experimenting with various feature comparison techniques to produce methods of video breakdown and identification. Video files were processed and converted into still images. The relevant frames from the video files were identified and sorted by scenes and shots. These images become a representative collection stored in the database. As they are identified and sorted, the images are processed to generate features for a video. The features collection provides the basis for comparison, recognition, and identification.

This document is currently not available here.

Share

COinS