| 摘要 |
PURPOSE:To improve efficiency by performing the quantization of a part where a quantization value to be encoded coincides with a predicted value with accuracy of (n+1) bits, and that of other part with accuracy of (n) bits. CONSTITUTION:A signal from an input part 1 is linearly quantized to (n) bits by a first quantizer 2, and an obtained quantization value is further quantized to the (n) bits by a second quantizer 4 by using the predicted value from a predictor 3, and is outputted to an output part 6. Simultaneously, the output of the quantizer 4 is reversely quantized at a reverse quantizer 5 by the predicted value, and is inputted to the predictor 3. In such a way, a quantization error is hidden, and is reduced to 1/2 in a visual sense, which realizes prediction with high efficiency. |
