An object contour extraction using local structure information captured by a spotlight filter

In this study, a spotlight filter for capturing the local structure information in tracing object's contour is proposed for object contour extraction. By searching the possible contour path between the initial source and destination contour points, the spotlight filter is able to capture the local structure information along the path, filter out the edge candidate pixels with weak evidence of being on contour and extract the correct contour. The spotlight filtering can then be operated in parallel on all pairs of initial source and destination contour points of an object to achieve a closed contour of the object. The proposed spotlight filter is tested on a synthetic object image with noise and a real medical image and better results are obtained in terms of accuracy and computational efficiency by comparing with other contour following methods.