Enhancing and Scaling PSSIM Calculator for Multiple Color Images

Members:Kevin W. Shay


Project Homepage:
Click here to view the project homepage

The p-value based structural similarity measure (PSSIM) was created for use in content-based image retrieval (CBIR) applications. It simulates the early stages of the human vision system and is more accurate than other image similarity metrics. The PSSIM divides the greyscale images being compared into blocks that are compared to an error block created by taking the difference in pixel values of corresponding blocks. The complexity of the calculation warrants parallelization of the algorithm. Enhancing the greyscale PSSIM calculator to measure multiple color images and determining the best way to parallelize it are the foci of this project. OpenMp, CUDA and OpenCL are used to create three implementations of a parallel PSSIM calculator. A graphical user interface was created to test the different versions and the results are evaluated and disclosed in a professional research paper. The GUI testbed and the final results may inspire and aid further research of the PSSIM.