RGB-Depth Image Based Human Detection Using Viola-Jones and Chan-Vese Active Contour Segmentation

Human detection refers to the process of detecting human region from an image or from video frames. Most of the recent advanced human detection systems use the segmentation scheme by incorporating the depth information of the scene. In such systems, the scene gets captured by a RGB-D camera and the candidate area is segmented by setting an appropriate depth threshold for the captured depth images. In practice, depth data obtained from this depth analysis having critical problems, such as optical noise, absence of depth information for certain regions like hair area, and unmatched boundaries. The proposed approach mainly focus on restoring the actual edge information and hair area of the subject in the pre-segmented image by applying Viola Jones Algorithm for face area detection and Chan-Vese active contour detection for restoring hair and edge areas of the image over the detected face area. This final segmentation mask is used for segmenting the accurate human region from the original image with hair area and with boundaries similar to the ground truth. Experimental results prove the improvement in the visual quality of the segmented human area.