Direct feature detection on compressed images

We present a general approach to the development of adaptive image processing operators of variable shape and size that can be applied directly to non-uniformly sampled images. The approach is demonstrated for second order derivative operators using content-based compressed images.