| 摘要 |
PROBLEM TO BE SOLVED: To improve encoding efficiency by independently encoding/decoding the object of any arbitrary rectangular shape and a predictive error expressed in the rectangular shape for the unit of rectangular object. SOLUTION: At an initial mesh generation part 2, a decoded previous frame is divided into regular rectangular initial meshes (or objects). At an object deforming part 4, this mesh is deformed into any arbitrary rectangle. At a motion estimating part 6, the motion of each rectangular object toward the time base is estimated and at a motion compensating and frame predicting part 8, a predictive object is formed from the rectangular object of the previous frame by video warping based on the estimated motion vector. At a predictive error forming part 20, a predictive error signal is generated for the unit of each rectangle. Namely, the predictive error exists inside the arbitrary rectangular object and at an object reconstituting part 22, the predictive error limited to the rectangular shape is transformed to the regular rectangular object in the size of K×K. At a predictive error encoding part 24, transforming encoding is performed to the error signal limited to the regular rectangular object while using the two-dimensional DCT of K×K. |
