Image encoding device and image encoding method

摘要

Provided are a highly efficient image encoding device which is achieved by quantization of each sub-block in accordance with visual characteristics, and a quantization method that uses the image encoding device. An input image is divided into macro-blocks having a prescribed first size, each of the divided macro-blocks is divided into sub-blocks having a prescribed second size, and each of the sub-blocks is encoded using a same or different quantization parameter for each of the sub-blocks. As a result, by executing quantization for each of the sub-blocks in accordance with the visual characteristics, highly efficient image encoding and image decoding devices can be provided.

主权项

1. An image encoding device comprising:
a MB division unit configured to divide an input image into macroblocks each having a first size; an evaluation unit configured to divide each of the macroblocks into sub-blocks each having a second size and evaluate a degradation cost in each of the sub-blocks; a determination unit configured to determine a quantization step size for an image of each of the sub-blocks based on the degradation cost evaluated by the evaluation unit; a quantization unit configured to quantize the image based on the quantization step size determined by the determination unit; and an encoding unit configured to encode the quantized image and information including the quantization step size for each of the sub-blocks in a syntax, wherein the determination unit determines the quantization step size by determining whether or not a Δ value is applied to a reference quantization step size with respect to each of the sub-blocks belonging to each macroblock, wherein the encoding unit adds a first code indicating which one of the sub-blocks in each of the macroblocks has the Δ value to the syntax, wherein the encoding unit adds a second code indicating the Δ value to a syntax in a picture header, and wherein the encoding unit selectively uses, as the first code, one of a 3-bits-code indicating eight patterns for the sub-blocks belonging to each macroblock and a 4-bits-code indicating fourteen patterns for the sub-blocks belonging to each macroblock and adds a third code indicating whether the first code is the 3-bits-code or the 4-bits-code to the syntax in the picture header.