CVPIC Compressed Domain Image Retrieval by Colour and Shape

Image retrieval and image compression have been pursued separately in the past. Only little research has been conducted on a synthesis of the two by allowing image retrieval to be performed directly in the compressed domain of images without the need to decode them first. In this paper we introduce a novel approach that provides such midstream content access [6]. Our work is based on the Colour Visual Pattern Image Coding (CVPIC) technique which represents a compression algorithm where the data in compressed form is directly visually meaningful. We present a compressed domain retrieval algorithm based on CVPIC that efficiently encapsulates the image content by colour and shape features. Retrieval results on the UCID dataset show good retrieval performance, outperforming methods such as colour histograms, colour coherence vectors, and colour correlograms.