No-Reference Quality Metrics for Image Decolorization

Evaluating the visual quality of decolorized images is challenging, as existing metrics such as CCPR and E-score depend on parameters that may vary across different methods. In this study, we propose novel no-reference quality metrics for image decolorization that are non-parametric, robust, and perceptually relevant. Our main contributions are: 1. We develop TIA and WTIA quality metrics that measure the preservation of salient image regions after decolorization. 2. We propose an image-dependent optimal decolorization method that uses TIA/WTIA metrics to adjust the decolorization parameters. 3. We conduct extensive experiments to show that our method produces better-decolorized images than state-of-the-art methods and that our metrics have a high correlation with subjective ratings from human observers.