Browse Teach Curate

Image Quality Assessment for Image Harmonization in Computer Vision

Predicting human visual preference for harmonized composite images, where a pasted foreground has been color- and light-adjusted to match a background. Built on a dedicated database of harmonized images with human preference scores and a large-multimodal-model predictor sensitive to the minor inconsistencies generic IQA metrics miss.

Data Text

HarmonyIQA: Pioneering Benchmark and Model for Image Harmonization Quality Assessment Zitong Xu

Image composition pastes a foreground object cut from one image onto another; image harmonization algorithms (IHAs) then adjust the foreground's color and lighting to match the background. General im…