Adaptive Image Encoding Using Region-Specific Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image encoding systems, such as JPEG, degrade image quality due to excessive reduction in data size, particularly in important regions like faces, leading to increased encoded data size, which is a concern for network communication efficiency.
Innovation Solution
An image processing apparatus that extracts object regions, selects important regions based on feature quantities, and applies different quantization step values for encoding, using a smaller step value for important regions to minimize quality degradation and reduce encoded data size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If lossy compression is applied to reduce data size, then encoded data size is reduced, but image quality degrades particularly in important regions
Solution Approach 1:
The patent applies different quantization step values to different regions of the image based on their importance. Important regions (such as faces or objects of interest) use a smaller quantization step value to preserve quality, while less important regions use a larger quantization step value to reduce data size. This local differentiation resolves the contradiction by maintaining quality where needed while compressing overall data size.
Solution Approach 2:
The patent dynamically changes the quantization step value parameter based on region importance. By adjusting this encoding parameter adaptively across different image regions rather than using a uniform value, the system achieves both quality preservation in critical areas and efficient compression in non-critical areas, resolving the trade-off between data size and image quality.
2Manufacturing precision
If high encoding amount is assigned to detected regions to maintain quality, then image quality is preserved, but encoded data size increases significantly
Solution Approach 1:
The patent selectively applies high encoding amounts (small quantization step values) only to important regions detected in the image, while applying lower encoding amounts (larger quantization step values) to the remaining regions. This localized approach preserves quality where it matters most while minimizing overall data size, resolving the contradiction between quality preservation and data compression.
Solution Approach 2:
The patent applies full encoding quality (higher encoding amount) only partially to important regions rather than uniformly across the entire image. By using excessive encoding quality only where necessary and adequate encoding elsewhere, the system achieves quality preservation in critical areas without the excessive data size increase that would result from applying high encoding throughout the entire image.
Data Source
AI summary
In an image processing apparatus, an object region is extracted from image data that has been input, and an important region where degradation in image quality is to be reduced is selected with use of a feature quantity of image data of the object region or a peripheral region in a periphery of the object region. Among portions of the input image data, an image data portion outside of the important region is compressed and encoded with use of a first quantization step value, and an image data portion of the important region is compressed and encoded with use of a second quantization step value that is smaller than the first quantization step value.