Compressed Domain Texture Retrieval Based on I-Frame Coding in H.264

In this paper a new texture descriptor is introduced for visual data. The proposed descriptor is based on the information of I-frame coding in H.264 video coding standard. In I-frame coding different prediction modes can be used for predicting a block from upper or left side neighboring pixels. The selected prediction mode for a block in an I-frame indicates the way that the block depends on the neighboring parts. In this paper, we use the prediction mode histogram of a compressed I-frame picture, as a texture descriptor for image or video retrieval. Simulation results comparing the proposed method with Gabor filter show that the proposed method is an efficient texture descriptor in compressed domain.