<p>This study develops a novel algorithm for clustering point data based on an improved self-updating function. In this algorithm, both the number of clusters and the elements within each cluster are determined automatically, relying solely on the variability of the data rather than on subjective user judgment. The algorithm is described in detail, including its implementation, computational aspects, and convergence analysis. A key motivation of this work is its application to image segmentation, where features are jointly extracted from the RGB color space and a deep learning model. Experiments on images with diverse characteristics show that the proposed segmentation method produces accurate and stable results, outperforming several popular algorithms in terms of visual quality and quantitative evaluation metrics. The algorithm can be readily applied to real data using programs developed in MATLAB software.</p>