Novel variance based approach to improving JPEG decoding

For JPEG image compression, there exists a well known approach to optimal reconstruction of quantized AC DCT coefficients in the JPEG decoder. However, the Laplacian distribution parameter must be firstly determined. To solve this problem, this paper proposes a practical novel variance based approach to estimate the Laplacian parameter based on the quantized DCT coefficients, which are already available in the decoder side. Both the theoretical analysis and extensive experimental results demonstrate that the proposed approach can achieve almost the best decoded image quality improvement when compared with other approaches reported in the literature. More importantly, the proposed approach provides a more practical and efficient manner to estimate the Laplacian parameter, and is hence more suitable for real-world applications. © 2005 IEEE.