Enhancing and Scaling a PSSIM Calculator
for Multiple Color Images
by Kevin Shay
sponsored by Dr. Venekat Gudivada

Introduction

P-value based structural similarity measurement (PSSIM) is used in content-based image retrieval (CBIR) applications. Any application involving anomaly detection or automated inspection can benefit from using the PSSIM to compare geryscale images. The PSSIM was developed by Haiyan Wang, Diego Maldonado and Sharad Silwal and implemented using R and Mathlab. The poor performance of this implementation warranted a translation into C++ and parallelization, initially using OpenMp. My PSSIM calculator is the culmination of these improvements. This research project further enhances the PSSIM calculator, enabling it to process many color images simultaneously. It will compare the performance of the PSSIM calculator implemented serially and parallelized using OpenMp, CUDA and OpenCL. A GUI will be created for the new PSSIM calculator to make it easier to test and apply to real world scenarios. Once completed, the performance of each implementation will be tested and the project details and results will be presented in a formal research paper.

Motivation

This project was chosen to demonstrate concepts learned while pursuing a computer science degree at Marshall University. Specifically, parallel programming, software development and testing, technical writing and an understanding of mathematical concepts are required to complete it.

Documentation

Vision and Scope

Software Requirements Specififcation

Research Paper

Research Paper

Prototype

Prototype.zip

News and Updates

I will be presenting my initial findings of parallelizing the PSSIM at the International Conference on Advances in Interdisciplinary Statistics and Combinatorics October 7 in Greensborogh, NC


Vision and scope document has been posted


SRS document Posted


Prototype finalized


Research paper posted