QUANTIZATION APPARATUS, QUANTIZATION METHOD, QUANTIZATION PROGRAM, AND RECORDING MEDIUM

摘要

With a simple configuration, the rounding accuracy upon quantization is dynamically changed so as to effectively suppress the encoding noise and prevent image deterioration. An image signal is subjected to DCT and a DCT coefficient obtained by this is input to a quantization circuit (30), quantized by a quantization matrix, and supplied to a multiplier (33) via an absolute value circuit. A quantization scale value in accordance with the generated code amount in the entire screen output from a rate control block (39) is inversed and supplied to the multiplier (33). Simultaneously with this, in an M setting block (38), a value M as a parameter controlling the rounding accuracy is set in accordance with the quantization scale value. In an adder (34), the value M is added to the output from the multiplier (33), and in a 5-bit shift circuit (35), rounding is performed by discarding the decimal portion. Here, as the quantization scale value increases, the quantization rounding accuracy is changed in the discarding direction. Thus, the DCT coefficient entered is qunatized without increasing its amplitude, thereby suppressing the encoding noise.