Study on Deriving the Optimal Method for Obtaining Optical Flow by Region

The tongue reflects the physiological and pathological state of the internal organs, and recent studies in both Oriental medicine and modern medicine have investigated its potential as an indicator of the health status of the tongue. This study aims to analyze tongue movement by comparing segmented and unsegmented tongue regions in sequential images. Four algorithmic approaches were evaluated: (1) histogram followed by Farnebäck optical flow, (2) tongue region segmentation followed by histogram equalization and Farnebäck optical flow, (3) histogram equalization followed by global motion aggregation (GMA) model, and (4) tongue region segmentation followed by histogram equalization and GMA analysis. Results showed that no single approach consistently outperformed the others. Each method showed different strengths depending on the analyzed region such as tongue tip, inner, or outer region. Therefore, the author applied the Farnebäck method and the GMA model to the tongue regions segmented from the tongue inner side and the tongue tip respectively to obtain optimized results. In future studies, we will focus on optimizing preprocessing and postprocessing techniques, as well as integrating region adaptation algorithms to improve the accuracy of motion analysis for disease prediction.