DCT quantization noise in compressed images

• Published in: IEEE transactions on circuits and systems for video technology Vol. 15; no. 1; pp. 27 - 38
• Main Authors: Robertson, M.A., Stevenson, R.L.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.01.2005; Institute of Electrical and Electronics Engineers
• Subjects: Applied sciences; Compressed image post-processing; DCT quantization noise; discrete cosine transform (DCT) coefficient distributions; Discrete cosine transforms; Error correction; Exact sciences and technology; Filters; Image coding; Image processing; Image representation; Image restoration; Information, signal and communications theory; Laboratories; Quantization; Sampling, quantization; Signal analysis; Signal and communications theory; Signal processing; Sun; Telecommunications and information theory; Quantization noise; Image processing; Compressed image post-processing; Noise reduction; Statistical model; Image compression; Image restoration; Discrete cosine transforms; discrete cosine transform (DCT) coefficient distributions; DCT quantization noise
• ISSN: 1051-8215; 1558-2205
• DOI: 10.1109/TCSVT.2004.839995

In lossy image compression schemes utilizing the discrete cosine transform (DCT), quantization of the DCT coefficients introduces error in the image representation and a loss of signal information. At high compression ratios, this introduced error produces visually undesirable compression artifacts that can dramatically lower the perceived quality of a particular image. This paper provides a spatial domain model of the quantization error based on a statistical noise model of the error introduced when quantizing the DCT coefficients. The resulting theoretically derived spatial domain quantization noise model shows that in general the compression noise in the spatial domain is both correlated and spatially varying. This provides some justification to many of the ad hoc artifact removal filters that have been proposed. More importantly, the proposed noise model can be incorporated in a post-processing algorithm that correctly incorporates the spatial correction of the quantizer error. Experimental results demonstrate the effectiveness of this approach.