Powered by OPF
OPF WIKI STATIC ARCHIVE
2,681 pages · 153 spaces · 776 tags · 4,025 history records · 96.2% of the original wiki recovered
Archived copy. This page was recovered from the Internet Archive snapshot of /pages/viewpage.action?pageId=6062236 taken on 2022-12-02. The original wiki at wiki.opf-labs.org is being decommissioned.

Identifying rotated, duplicate images using pHash

Added by Gareth Knight · last edited by Gareth Knight · on Jun 14, 2011
One line summary Identifying duplicate images rotated during post-processing
Detailed description Several tools, including XCL Extractor/Comparator, ImageJ and pHash were briefly explored to determine if an image was a rotated copy of another image. The solution champion discovered that pHash was able to compare two images and make an assessment on the degree of similarity. Two images are compared and evaluated according to their level of similarity. If the images fit within a specified threshold (28% in the website implementation) they are declared as likely to be similar.
Solution champion Peri
Git link  
Evaluation Testing was performed on the pHash web site, comparing original images with those rotated within the defined parameters (0-5%). pHash correctly identified them as being similar. Different images that share some common headings have been compared and were (correctly) identified as different. Further work is necessary to develop a software tool that integrates the pHash library.
Tool (link) pHash