AcademySoftwareFoundation
[BUG] idiff's perceptual comparison passes images with different dimensions and contents
Describe the bug
Running idiff's perceptual difference test on two images with different dimension and obvious content differences passes the test.
These are the images being compared:
OpenImageIO version and dependencies
OIIO 3.1.7.0 | MacOS/x86_64 Build compiler: Apple clang 16.0 | C++17/201703 HW features enabled at build: sse2,sse3,ssse3,sse41,sse42 No CUDA support (disabled / unavailable at build time) Dependencies: BZip2 1.0.8, FFmpeg 8.0, fmt 12.1.0, Freetype 2.14.1, GIF 5.2.2, Imath 3.2.2, JXL 0.11.1, Libheif 1.20.2, libjpeg-turbo 3.1.2, LibRaw 0.21.4, libuhdr NONE, OpenColorIO 2.5.0, OpenEXR 3.4.2, openjph 0.24.5, OpenVDB NONE, PNG 1.6.50, pugixml 1.15, pybind11 3.0.1, Python3 3.14.0, Robinmap 1.4.0, TBB 2022.3.0, TIFF 4.7.1, WebP 1.6.0, ZLIB 1.3.1
(Installed and upgraded to the latest version of OpenImageIO with homebrew)
To Reproduce
Steps to reproduce the behavior:
- Download and extract the images from this ZIP file: images.zip
- In the directory with the two images, run
idiff -p -abs -o diff.jpg nav_buttons_new.jpg nav_buttons.jpg
It outputs
Comparing "nav_buttons_new.jpg" and "nav_buttons.jpg" PASS
even though the images are clearly very different. This is the diff that is generates:
Passing -od to the command makes it not generate any diff because the comparison passes.
The expected result would be that the comparison fails.
@keiwando how this PR might fix it, can you verify it please https://github.com/AcademySoftwareFoundation/OpenImageIO/pull/4976
It's instructive to note that idiff without the -p correctly identifies the images as different. That makes me think that the bug is specific to the perceptual comparison, and not to idiff overall. I suspect you should be searching in ImageBufAlgo::compare_Yee() to discover where things are going awry when the two images do not have the same pixel window.
