Perceptual Image Diff comparison

Skip to end of metadata
Go to start of metadata
One line summary A wrapper around Perceptual Diff Image Comparison tool to compare images, possible issues may be detected if the results are not sufficiently close.
Detailed description Simple wrapper script which walks through a folder and compares each image using Perceptual Image Diff which  emulates the way a human looks at an image.
It uses a treshold of number of pixels that are allowed to be different.

Consult the Perceptual Image Diff webpage for more in-depth information about this technology.

The tool reports back:
- if images are (almost) the same and reports back the number of different pixels (PASS)
- if images are not the same (FAIL)
- if an error occured (ERROR)

Solution champion Maurice de Rooij
Git link
Group Evaluation Notes
  • Several approaches explored, some drawbacks/obstacles encountered
  • Potential to explore the configuration and trial with larger sample sets
  • WSQ approach may yield best results but requires more investment in building a usable tool
  • First version of wrapper is written in PHP. Maurice will rewrite this in python, since it is more industry-standard.
  • Mette and Maurice will do some further testing
  • Could also do further use/investigation at next AQuA event
  • Fingerprinting is very memory hungry, Maurice recommends Linux rather than Windows
  • De-duplication application requires a lot of comparisons, e.g.: comparing 100 files takes 56 comparisons.
  • QA application more suited for this reason
Detailed Evaluation
Possibly a useful tool for first 'quick' scan of duplicates in collections of scans and images.
It is quiet slow on Windows and eating up a lot memory when comparing images which are bigger than 100kB.
CPU is probably not an issue. On Windows it used max 50 percent of CPU power, as reported by taskmanager. This might be a false number, needs further investigation.
Tool (link) Perceptual Image Difference utility
aqua aqua Delete
solution solution Delete
de-duplication de-duplication Delete
quality_assurance quality_assurance Delete
Enter labels to add to this page:
Please wait 
Looking for a label? Just start typing.