Video Encoding Specific Area Prediction and Quality Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding techniques face challenges in accurately detecting specific areas, such as human faces, within moving images at high frame rates, leading to reduced image quality and propagation of quantization errors due to low prediction accuracy.
Innovation Solution
An image processing apparatus and method that sets a target image with a specific area and a non-target image, predicts the specific area in the non-target image using interpolation or extrapolation, and encodes the images so that the specific area has higher quality than other areas, without referencing the specific area from another frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specific area detection processing is executed for all frames at high frame rate, then prediction accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent applies preliminary action by detecting specific areas in advance (e.g., in key frames) and using these pre-detected results to predict specific areas in subsequent frames without performing full detection processing. This allows the system to maintain high prediction accuracy while reducing the computational burden and processing time required for each frame.
2Productivity
If motion vector prediction is used to predict specific areas, then processing time is reduced, but prediction accuracy decreases due to minimization of pixel value differences
Solution Approach 1:
The patent introduces an intermediary approach by using detected specific areas from key frames as reference points to predict specific areas in intermediate frames. Instead of relying solely on motion vectors that minimize pixel differences, the system uses the intermediary information from accurately detected specific areas to maintain prediction accuracy while processing all frames at the encoding frame rate.
Solution Approach 2:
The patent applies local quality by differentiating the processing approach for different regions. Specifically, it performs accurate specific area detection only in key frames where it is most needed, while using prediction based on previously detected specific areas for other frames. This localized application of high-quality detection maintains overall prediction accuracy while reducing total processing requirements.
3Quantity of substance
If quantization values are increased for non-specific areas, then code amount is suppressed, but image quality of predicted specific areas may drop due to error propagation
Solution Approach 1:
The patent applies preliminary action by pre-detecting specific areas in key frames before encoding, and using these pre-detected specific areas as references for prediction in subsequent frames. This preliminary detection ensures that the specific areas are correctly identified and encoded with appropriate quality, preventing quantization errors from propagating to subsequent frames while still allowing aggressive quantization in non-specific areas.
Data Source
AI summary
An image of a prescribed frame of images of respective frames is set as a target image, and an area including a prescribed pattern is detected from the target image as a specific area. An image other than the target image is set as a non-target image, and the specific area in the non-target image is predicted. The images of the respective frames are encoded so that the specific area is encoded to have higher image quality than an area other than the specific area. In encoding, the images of the respective frames are encoded so that the specific area in the non-target image is not referred to from another frame.