| 摘要 |
<p>PURPOSE:To prevent the accumulation of a quantized noise by encoding a reference sample value by a quantization bit number (n) with a negligible quantization error and a quantization bit number (m) which is smaller that said number (n) and prodduced m the basis of a difference related to said reference sample value. CONSTITUTION:Among three sample values A3i-1, A3i and A3i+1, bit are encoded by n-bits with the sample value A3i denoted by a circle as a reference sample value. Here the quantization bit number (n) is a value with a negligible qantization error. In terms of the remaining two sample values A3i-1 and A3i+1, two difference data B3i-1 and B3i+1 which are denoted by a cross and given according to the equation in accordance with the difference from the reference sample value A3i are encoded by the bit number m(<n). Where n=8 and m=4, the reference data A3i is encoded by eight bits, and the difference data B3i-1 and B3i+1 are encoded by four bits. Compared with a conventional system encoding all sample values by eight bits, bits can be compressed to 2/3.</p> |
