| 摘要 |
PURPOSE: To quantize a residual signal with high accuracy. CONSTITUTION: The residual signal generated by predictive encoding is divided for each block by a block circuit 2. Thus, the level distribution of the residual signals in the respective blocks is concentrated in the narrow block in comparison with the level distribution of entire one picture. Reference values DR and MIN of the residual signal for the unit of this block are detected and quantizing step width Δ and shift data ΔS are decided from these reference values and a number N of quantizing bits. At a quantizing circuit 6, the residual signal is quantized by the Δ. Besides, even when the level '0' of the residual signal is set on the decoding side, the residual signal is shifted and quantized later so that it can be provided as '0'. A flag FLG showing the quantized value at the level '0' and the Δ are transmitted together with a quantized value Q as side information. |
