An efficient image retrieval method under dithered-based quantization scheme

Recent research works have shown the impact of the quantization techniques on the performance of standard image retrieval systems when datasets are compressed in a lossy mode. In this work, we propose to design an efficient retrieval method well adapted to wavelet-based compressed images. Our objective is to recover features of the original image (herein the moments of the unquantized subbands) directly from the quantized coefficients. To this end, we propose to apply a dithered quantization technique satisfying some specific conditions. Then, the estimated moments of the wavelet subbands are used in an appropriate way to construct the feature vectors of the database images. Experimental results show the interest of the proposed image retrieval method compared to the state-of-the-art ones.