Error analysis and consistency improvement for human skin color videos

Skin color video measurements are very important in remote and non-invasive physiological monitoring, and its consistency always affects the measurement accuracy of physiological parameters like heart rate. An image-stabilization method was presented based on Scale Invariant Feature Transform (SIFT) and Wavelet Transform(WT), and was applied to improve the consistency of skin videos image. The video image error was analyzed and an error mathematical model was built. Based on that, the matched pair-points of skin video images was extracted using SIFT algorithm, and on the global motion parameter estimation was carried for the skin videos image to achieve the motion compensation. Finally, a statistics computation was used to evaluate the consistency effects. The results show that the proposed method is valid and correct for improving the consistency of skin videos image, and it lays a foundation for further physiological parameters calculation and studying the optical characteristics of the acupoints.