JPEG2000 ENCODING METHOD FOR REDUCING TILING ARTIFACTS

摘要

This paper proposes an effective JPEG2000 encoding method for reducing tiling artifacts, which is one of the biggest problems with the JPEG2000 encoder. It is generally thought that the main reason for the tiling artifacts is the symmetric extension of pixels. This paper shows that a more significant reason for the tiling artifacts at the middle or low bit rate is the differences of the quantization accuracy between tiles. This paper also proposes a method consisting of the two basic algorithms below for compensating the differences: (1) an algorithm that predicts whether tiling artifacts will occur at a tile boundary in the rate control process, and that locally improves quantization accuracy with the control of the post quantization, and (2) an algorithm that estimates the amount of information of each tile, and that allocates bits according to it in order to reduce the difference of quantization accuracy between tiles. This paper also proposes a method of reducing processing time, which is another serious problem of the JPEG2000 encoder. The method predicts the truncation points using the entropy of wavelet transform coefficients prior to the arithmetic coding. These encoding methods require no additional processing in the decoder. In the experiment, it was subjectively confirmed that tiling artifacts were greatly reduced and that the coding process was considerably accelerated.