Fast footprint contour extraction by curve evolution via level sets

A method based on spatial multi-scale analysis is put forward. Based on the energy projection distribution on rows and columns of an image, the image is divided into subimages with different size, which is adaptively adjusted according both the holistic features and the local features. By assigning the average of gray levels in each subimage as the gray level of the corresponding element in a new image, the original image is compressed. Curve evolution method via level sets is then applied to the newly compressed image, and the data size is reduced largely. In addition, the time of extracting contours is shortened substantially in turn and the practicability of the method is enhanced significantly. The proposed method is used to extract footprint contours and is shown experimentally to outperform the direct edge detection method as well as the curve evolution method basing on level sets in terms of computational cost and the detection veracity.