| 摘要 |
<p>PURPOSE:To reduce the number of bits used for an operation and to improve arithmetic accuracy by separating an input matrix into a direct current component and the other component, synthesizing results obtained with separately executing orthogonal inverse transformation and executing the orthgonal inverse transformation of the inputted matrix. CONSTITUTION:A discrete-cosine inverted input matrix A is separated into a direct current component matrix A-A deg. and the other component matrix A deg.. In calculating blocks 2 and 6, respective multipliers 21 and 61 multiply input signal fi by a coefficient Ci, and adders 22 and 62 obtains their sum of products SIGMACi.fi. The calculation result in a block 2 is sent to a range setting circuit 3, a variable range digit (m) is obtained by an expression I, and the range is determined from the digit (m) to m-k+1. The set range is shifted, a value is rounded, delivered to the block 6, the inverse transformation T<-1>A deg. is obtained, and sent to an adder 7. On the other hand, the matrix A-A deg. of the direct current component is inputted to a divider 8, the inverse transformation T<-1> (A-A deg.) is obtained, sent to the adder 7, and added with the inverse transfor mation. T<-1>A deg.. Consequently, the inverse transformation T<-1>A of the input matrix A can be obtained by the expression I.</p> |
