Image-based facial pore detection and visualization in skin health evaluation

The role of health information visualization and visual analytics processes is gaining importance. The facial pore feature is one of the crucial indicators of skin health evaluation. However, pores are tiny, which is difficult to detect and analyze based on the digital picture. In this paper, we aim to visualize facial pores by fabricating data to display their different roughness levels. First, we presented an image-based facial pore detection algorithm that combines the characteristics of skin pigment distribution and optimal scale. Second, based on the pore detection result, we proposed a P-DBSCAN (Pore density-based spatial clustering of applications with noise) algorithm that integrates pore characteristics. Because the local saliences of pores and interferences are different at the biological perspective, the interferences can be considered as noisy data in the scheme of well-known DBSCAN algorithm. As a result, the proposed algorithm determines two essential thresholds for facial pore detection and visualization, and makes it possible to improve the detection accuracy and accomplish visualization. On that basis, an index to objectively evaluate the roughness of skin pores was established by using the optimal scales in SIFT. The experiment results suggest improved accuracy of pore detection, and the facial pore visualization presents pore information directly and efficiently.