An experimental study of object detection in the wavelet domain

Most image and video data is stored or transmitted after compression for efficiency. Processes like pattern detection and localization typically include the extra expense of decompressing the data since most image and video processing requires access to the original pixel value in the spatial domain. It may, however, be more computationally efficient if we directly process the compressed bit stream or operate in the intermediate transform domain without fully decompressing the data. Coders like SPHIT and JPEG 2000 are built around the wavelet transform, so in this paper we study detection and localization in the wavelet domain.