Multiscale CCM Noisy Color Image Segmentation Algorithm with Channel Weighting

Credal c-means clustering (CCM) is an improved version of evidence c-mean, widely used in data clustering and image segmentation. However, CCM underutilizes spatial information and fails to consider differences among channel characteristics, resulting in poor performance for color image segmentation. To address these limitations, we propose multiscale CCM noisy color image segmentation algorithm with channel weighting. The proposed algorithm incorporates the following key improvements. Firstly, a multi-scale local information factor is developed by taking into account the relevant spatial information at different scales, thereby enhance the algorithm's ability to handle noisy images. Secondly, a penalty term is constructed and introduced into the objective function to regulate the formation of meta-clusters, ensuring more precise segmentation results. Lastly, a channel weighting strategy is designed to dynamically determine the weighting coefficients based on the contribution of different color channels to the segmentation results. Extensive experiments conducted on BSD500 datasets have demonstrated the strong performance of the proposed algorithm in segmenting noisy color images.