SINGULAR-VALUE EXPANSION ENCODING SYSTEM

摘要

PURPOSE: To provide the singular-value expansion encoding system which improves the compression efficiency by performing the encoding processing after reducing the redundancy of eigenvectors. ;CONSTITUTION: Picture data in block units is given from a picture input part 1 in a form of (g) and is sent to a singular vector operation part 2. This part 2 calculates an N×N-order discrete self-correlation matrix R_n and an average R of every N components to obtain eigenvectors and corresponding elgenvalues. Eigenvalues are arranged in the order of absolute value to obtain an eigenvector matrix A, and corresponding eigenvectors are rearranged in accordance with this order to obtain an eigenvector matrix Φ. A subtraction result E between the eigenvector matrix Φ and an average eigenvector preliminarily stored in a storage part 4 and the eigenvalue matrix Λ are sent to a quantization part 5. Since the average eigenvector ψ is subtracted from the eigenvector matrix Φ, the difference E becomes a very small value (or 0), and the compression ratio is considerably improved.;COPYRIGHT: (C)1994,JPO&Japio