Content based image retrieval utilizing HEVC intra coding features

Abstract

In this paper, a content-based image retrieval technique applied to compression domain for the newest video coding standard, high efficiency video coding (HEVC), is developed. We propose to extract the features from intra frames, which are only partially decoded. The directions of intra mode are taken as a part of the feature when the corresponding sizes of prediction unit are divided into two groups. The second part of the feature is the histogram of the signal power of residuals. The experimental results show that the proposed method can reduce the resource consumption and achieve a good retrieval performance in Oxford 5K dataset.