Entropy-based indexing and retrieval system of medical images using FPGA

Date of Publication


Document Type

Master's Thesis

Degree Name

Master of Science in Electronics and Communications Engineering


Gokongwei College of Engineering


Electronics And Communications Engg

Thesis Adviser

Cesar A. Llorente

Defense Panel Chair

Roderick Y. Yap

Defense Panel Member

Edwin Sybingco
Reggie C. Gustilo
Jonathan R. Dungca


This research aims to improve the accuracy of a content-based image indexing and retrieval system (CBIR) and the efficiency of its implementation in FPGA. The histogram of the image is widely used as basis in computation of entropy for CBIR, however, different images having similar distribution on the histogram can be falsely detected as similar images. In the proposed system, the image is segmented into mxn blocks, entropy is then computed for each block based on its color histogram. This method aims to improve the accuracy of the image retrieval. The query image will undergo the same process to compute for its entropy. In retrieving images from the database, the similarity between two images is computed using Euclidean distance. Images that have least distances are included in the list of candidates for the query. The researcher has able to establish that 16x16 block partitions would yield the highest accuracy. Also, Thresholding was found to have improved the outcome of the query. Furthermore, the researcher has also optimized the limited size of the RAM of the FPGA by reducing the image resolution to 128x128, and the processing time was reduced by applying HSTRP and VSTRP block patterns with improved accuracy. Finally, the research has successfully developed and implemented an alternative way of computing for square root herein referred to as the SQUART method.

Abstract Format






Accession Number


Shelf Location

Archives, The Learning Commons, 12F Henry Sy Sr. Hall

Physical Description

1 computer disc ; 4 3/4 in.


Entropy; Diagnostic imaging; Imaging systems in medicine

This document is currently not available here.