A Dominant Color Extraction Method Based on Salient Object Detection
Chaonan Bao, Jian Hu, Yuxian Mo, Dawei Xiong · 2023
Dominant color extraction is essential for image color analysis in the field of image processing. However, traditional dominant color extraction methods are unable to accurately characterize the whole image because they only extract the colors that appear more frequently in an image and ignore small regions of prominent colors. In this paper, a novel dominant color extraction method is proposed. Firstly, the location information of the prominent colors in small regions of an image is obtained by using salient object detection technology. Secondly, the most representative colors in the whole image and the region are accurately extracted as the candidate dominant colors by using a clustering-based dominant color extraction method. Finally, the colors with small differences in the candidate dominant colors are eliminated by using a color similarity deduplication strategy to obtain the final dominant colors. Experiments show that our method not only extracts the colors that appear more frequently in an image but also captures the prominent colors in small regions, which can accurately characterize the colors of the whole image and improve the effectiveness of dominant color extraction.