A method of sequentially predicting DCT coefficients using gradient information of an input signal and its application to image compression (with K. Ashizawa, J. Ogawa, and K. Yamatani), Urban Science Studies, vol.14, pp.63-70, 2009 (in Japanese).


The JPEG standard is the most popular still image compression technique based on the discrete cosine transform (DCT). The JPEG standard achieves a high compression ratio with little perceivable loss in image quality by the quantization of the DCT coefficients and the entropy coding. The quantization of the low frequency components, however, causes the blocking artifacts. The quantization of the high frequency components also influences the representation of the texture and the object contour. In this paper, we propose a new multiple prediction scheme of AC coefficients for the DCT. Our purpose is to improve the above situation based on the AC prediction scheme for the JPEG standard. Our mathematical concept is to match the prediction in terms of the gradients at the several points equally arranged on the DCT block. We numerically compared our method with some of the other AC prediction methods previously proposed using the eight standard images.

Get the full paper: PDF file.

Please email me if you have any comments or questions!
Go back to Naoki's Publication Page