| 摘要 |
<p>PURPOSE:To preclude an overflow and eliminate an error in recovery by minimizing the numbers of bits of a conversion coefficient, a storage means, and an output under specific conditions when rounding is performed by the storage means of the two-dimensional converter and then discarding is performed. CONSTITUTION:When the rounding is performed by the storage means 11 of the two-dimensional linear converter, 1 is set to 1 and when the discarding is performed, 1 is set to 2; when neither one is performed, 1 is set to 0. Then, 1 is set to 1 when the rounding is performed with the output of the two-dimensional linear converter, and to 2 when the discarding is performed. Further, 3 is set to 1 when the rounding is performed by the storage means 18 of the two-dimensional linear reconverted ER to 2 when the discarding is performed, and to 0 when neither one is performed. Then, the number of bits of the approximate matrix of conversion coefficients and the numbers of bits of the integer part and decimal part, i.e., I1, D1, I2, D2, I3, and D3 are minimized satisfying equations I, II, III, and IV. Here, T1 and T2 are N.N conversion coefficient matrixes.</p> |
