Image Encoding Apparatus Adaptive Quantization Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding schemes, such as H.264, face increased computational costs and bit requirements when controlling quantization values across multiple blocks in a frame, as they need to consider the number of blocks in each class to predict bit amounts, leading to inefficient code allocation and image quality management.
Innovation Solution
An image encoding apparatus that calculates block evaluation values based on visual characteristics, determines block classes, and adjusts quantization parameters based on the number of blocks in each class, allowing for adaptive control of quantization values to balance image quality and code allocation without counting all blocks in a frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the number of blocks belonging to all classes in each frame is counted to predict the generated bit amount, then the quantization value control can be improved, but the computational cost and bit requirements increase
Solution Approach 1:
The patent segments the frame into multiple classes based on visual characteristics (e.g., skin regions, text regions, general regions) and processes each class separately. Instead of counting all blocks uniformly, the system counts blocks per class and uses class-specific quantization parameters. This segmentation allows precise control where needed while reducing overall computational complexity by treating different regions differently.
Solution Approach 2:
The patent applies local quality by assigning different quantization parameters to different classes of blocks based on their visual importance. Skin regions receive lower quantization values (higher quality), text regions receive moderate values, and general regions receive higher values (lower quality). This local differentiation improves overall image quality control while reducing the need to uniformly process all blocks with high computational cost.
2Manufacturing precision
If a lower quantization value is assigned to blocks belonging to a class for remarkable degradation of image quality, then image quality degradation is inhibited, but the amount of code increases
Solution Approach 1:
The patent applies local quality by assigning different quantization parameters to different classes of blocks based on their visual importance. Skin regions receive lower quantization values (higher quality), text regions receive moderate values, and general regions receive higher values (lower quality). This local differentiation improves overall image quality control while reducing the need to uniformly process all blocks with high computational cost.
Solution Approach 2:
The patent dynamically changes the quantization parameter based on the class of each block. By adjusting the quantization value parameter according to visual importance (lower for skin regions, higher for general regions), the system optimizes the balance between image quality and code amount, ensuring critical regions maintain quality while less critical regions use fewer bits.
Data Source
AI summary
An image encoding apparatus which encodes an image includes a first acquiring unit configured to acquire an evaluation value for a block being a division of the image, a determining unit configured to determine a class to which the block belongs to among a plurality of classes based on the evaluation value acquired by the first acquiring unit, a second acquiring unit configured to acquire count information describing the number of blocks belonging to a predetermined class of the plurality of classes, and a setting unit configured to set a quantization parameter usable for encoding a block belonging to a different class from the predetermined class based on the count information acquired by the second acquiring unit.