Skip to content

A high performing paralleled image searching program

Notifications You must be signed in to change notification settings

namnbk/Image-Search

Repository files navigation

A High Performing Image Searching Program

Identify and mark (with a red-box) occurrences of a given sub-image in a larger image

Overview

  • The image search program must be designed to handle searching for matches using an image “mask”. An image mask is a black-and-white image that essentially represents a pattern to search for

  • Once a region of the image has been matched, it must not be included as part of any other matches. For each matched region, draw a red box around the region.

  • Design, implement, and validate suitable parallelization approach for this problem using OpenMP.

Environment

On the Ohio Supercomputing Center Pfizer cluster

Component Details
CPU Model Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
CPU/Core Speed 2.40GHz
RAM 200GB
Operating system used Linux 3.10.0-1160.95.1.el7.x86_64 #1 SMP x86_64 x86_64 x86_64 GNU/Linux
Name and version of C++ compiler gcc version 8.4.0 (GCC )
Name and version of other non-standard software tools & components OpenMP

Result 1

Large Image Mask Detected Result
Mamogram Mamogram Masked Mamogram Detected

Performance Statistics

#Threads User Time (s) Elapsed Time (s) %CPU
1 67.54 67.73 99%
4 65.54 17.46 375%
8 68.92 09.30 740%

Result 2

Large Image Mask Detected Result
Mamogram Mamogram Masked Mamogram Detected

Performance Statistics

#Threads User Time (s) Elapsed Time (s) %CPU
1 48.54 48.85 99%
4 50.41 15.97 315%
8 51.21 10.10 507%

About

A high performing paralleled image searching program

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published