Adaptive image processing via snake filters

We propose an image enhancement and scaling technique using adaptive snakes that modify the image according to region area and perimeter. Snake filters consist of geometric active contours that eliminate image regions below the area scale and above the specified perimeter. The snake filters are efficiently implemented using front propagation via a fast marching method. We compare the snake filters to similar filters based on connected component analysis and area morphology. The paper demonstrates that snake filters improve upon the computational cost and the flexibility of area morphological filters such as the area open-close filter. Applications of object detection based on area and perimeter scales are given for biomedical images.