Mathematical Problems of Computer Science (Jun 2022)
A New Image Decolorization Evaluation Quality Metric
Abstract
Image decolorization, the process of color-to-gray conversion, plays a crucial role in single-channel processing, computer vision, digital printing, and monochrome visualization. This process induces new artifacts, the impact of which on visual quality has to be identified. While visual quality assessment has been the subject of many studies, there are still some open questions regarding new color-to-gray conversion quality metrics. For example, computer simulations show that the commonly used grayscale conversion quality metrics such as CCPR, CCFR, and E-score depend on parameters and may pick different best decolorization methods by changing the parameters. This paper proposes a new quality metric to evaluate image decolorization methods. It uses the human visual properties information and regression method. Experimental results also show (i) strong correlations between the presented image decolorization quality metric and the Mean Opinion Score (MOS), (ii) more robust than the existing quality metrics, and (iii) help to choose the best state-of-the-art decolorization methods using the presented metric and existing quality metrics.
Keywords