Towards key frame extraction in compressed domain

Following the launch of MPEG-7 towards content description and interoperability, we propose our design in directly extracting MPEG-7 dominant colour descriptors from MPEG compressed videos, which will provide useful tools for MPEG-7 applications to those compressed videos, where no metadata is available. By examining those dominant colour components extracted in I-frames, we further describe our design to extract key frames from those compressed MPEG videos, which enables fast search and browsing of the compressed video content without involving full decompression/IDCT. Experimental results delivered support our design, which are also included in this paper.