| 摘要 |
<p>An 8×8 integer transform method in video coding is disclosed by the present invention, which includes: receiving 8×8 image residual error data X output by a prediction module of a coder; selecting different integer transform bases [15, 15, 10, 3], [6, 7, 5, 1] to construct a transform matrix P; according to the transform matrix P, constructing a scaling matrix PF=ZT*Z; performing forward transform on the image residual error data X according to Y=PXPT, and adopting corresponding integer transform calculation method according to the different integer transform bases; and performing scaling processing on the 8×8 forward transformed data Y. The transform method of the present invention has excellent energy concentration performance, wherein the degree of energy concentration approaches the 8×8 Discrete Cosine Transform (DCT), thus greatly eliminating the correlation of video data in transform space. The transform method also has low calculation complexity, which can be realized by addition and shift and is easy to implement by hardware; and the transform method has high calculation precision, and thus there does not exist the problem that the forward transform does not match the inverse transform.</p> |
