Face Region Compression Ratio Control for Video Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video compression techniques fail to maintain high image quality for face regions while reducing bit rates effectively, as they often misidentify non-face regions as attention areas and do not adequately adjust compression ratios based on face area size, leading to degraded image quality in small-face scenarios.
Innovation Solution
A moving image processing device that includes a specific target detector to identify face regions, a video compressor to adjust compression ratios dynamically, and a compression ratio controller to reduce the compression ratio of face regions based on their area size within the image, ensuring adequate image quality for both large and small face areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a higher compression ratio is set for non-face regions to reduce bit rate, then bit rate reduction is achieved, but image quality in those regions deteriorates
Solution Approach 1:
The patent applies different compression ratios to different regions of the image based on whether they contain face portions. Face regions use a first compression ratio (lower compression, higher quality) while non-face regions use a second compression ratio (higher compression, lower quality). This local differentiation resolves the contradiction by allowing aggressive compression in non-critical areas while preserving quality in face areas.
2Manufacturing precision
If a lower compression ratio is set for face regions to maintain image quality, then image quality of face regions is improved, but bit rate increases
Solution Approach 1:
The patent selectively applies a lower compression ratio (first compression ratio) only to regions containing face portions, while applying a higher compression ratio (second compression ratio) to all other regions. This localized approach maintains face quality without unnecessarily increasing the bit rate for the entire image, thus resolving the contradiction between face quality and overall bit rate.
3Ease of operation
If conventional face detection based on color information is used, then detection simplicity is maintained, but detection accuracy deteriorates due to misidentification of non-face regions
Solution Approach 1:
The patent employs multiple detection methods with different reliability levels. It first uses simple color-based detection, then applies more sophisticated methods (pattern matching, neural networks, or other object detection techniques) to verify and correct detections. This dynamic, multi-stage approach maintains simplicity while improving accuracy by adaptively applying more complex methods only when needed.
4Device complexity
If uniform compression ratio is applied to entire image, then processing simplicity is maintained, but image quality of attention regions deteriorates
Solution Approach 1:
The patent divides the image into face regions and non-face regions, then applies different compression ratios to each region type. Face regions receive a first compression ratio optimized for quality, while non-face regions receive a second compression ratio optimized for bit rate reduction. This local differentiation resolves the contradiction by maintaining processing simplicity through automated region classification while achieving superior quality in attention regions.
Data Source
AI summary
A moving image processing device (200) includes a face detector (3) (specific target detector), a compression ratio controller (4), and a quantizer (9) (video compressor). The face detector (3) detects a person's face included in an input image and extracts a face region. The compression ratio controller (4) generates and outputs a compression control instruction for reducing the compression ratio of the face region extracted by the face detector (3) so that a degree of reduction in the compression ratio of the face region is small in the case where the ratio of the area of the face region to the entire area of the input image is relatively large and so that the degree of reduction in the compression ratio of the face region large in the case where the ratio of the area of the face region to the entire area of the input image is relatively small. The quantizer (9) performs a quantization processing of the face region according to the compression control instruction given from the compression ratio controller (4).