Impact of image processing pipeline on video-based heart rate estimation

Although previous video-based heart rate (HR) estimation methods operate on full-color images processed through an image processing pipeline, such pipelines may distort subtle blood volume pulse (BVP) components. In this paper, we analyze the impact of the image processing pipeline on video-based HR estimation. In addition, we introduce an BVP signal extraction method that utilizes raw data directly from the RGB image sensor. Based on medical knowledge of BVP signals, we can effectively extract the BVP signal from raw data, enabling us to improve HR estimation performance.