Video Coding Device Mode Switching and Screen Content Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in managing the transition from still-image coding to video coding, leading to increased information amounts and potential image deterioration, especially when dealing with screen content images that have low pixel correlation and are sensitive to format changes like YCbCr 4:4:4 to YCbCr 4:2:0 conversions.
Innovation Solution
A video coding device and method that dynamically switches between still-image coding and video coding based on image motion, using inter prediction coding with a decoded frame as a reference to minimize information increase, and employs specific filtering techniques to maintain image quality by treating the decoded frame of a coded still-image as a reference for the first frame after switching, thereby reducing the amount of information and image deterioration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If still-image coding is applied to screen content images, then compression efficiency is improved, but image quality deteriorates due to low pixel correlation and mosquito distortions
Solution Approach 1:
The patent applies different coding methods to different regions of the image. For screen content regions with low pixel correlation, it uses intra prediction coding with specific prediction modes (e.g., planar, angular) rather than DCT-based transformation. This local adaptation of coding strategy to match regional characteristics reduces mosquito distortions while maintaining compression efficiency.
Solution Approach 2:
The patent changes key coding parameters when detecting screen content images, including disabling transformation skipping, adjusting prediction mode selection, and modifying quantization parameters. These parameter adjustments are triggered by scene detection algorithms that identify screen content characteristics, allowing the encoder to optimize for screen content rather than natural images.
2Productivity
If DCT is applied to screen content images with low pixel correlation, then compression is achieved, but mosquito distortions occur in high-frequency components
Solution Approach 1:
The patent disables DCT transformation for screen content regions and instead uses prediction-error coding without transformation. This local suppression of DCT application in screen content areas eliminates the source of mosquito distortions while maintaining compression through efficient prediction and coding of residuals.
Solution Approach 2:
The patent converts the low pixel correlation characteristic of screen content (which causes problems with DCT) into a benefit by using prediction methods that exploit the structured nature of screen content, such as text and graphical elements, rather than relying on spatial frequency transformation.
3Object-affected harmful factors
If inter prediction coding is used for the first frame after switching from still-image coding, then information amount increases, but image quality is maintained
Solution Approach 1:
The patent performs preliminary preparation by storing the decoded still-image frame in a reference frame buffer before switching to video coding mode. This preliminary action enables immediate inter prediction coding without delay, using the still-image frame as a reference, thereby maintaining image quality while managing information flow efficiently.
Solution Approach 2:
The patent uses a reference frame buffer as an intermediary between still-image coding and video coding modes. This intermediary structure allows seamless transition and enables the first video frame to reference the still-image frame, reducing information requirements compared to pure intra prediction while maintaining quality.
4Productivity
If intra prediction coding is used for screen content images, then compression efficiency is improved, but coding complexity increases
Solution Approach 1:
The patent segments the coding process into distinct modes: screen content detection, intra prediction coding without transformation, and reference frame management. This segmentation allows each component to be optimized independently, reducing overall coding complexity while maintaining compression efficiency for screen content.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A storage unit (311) stores a decoded image of a coded image that is coded before a coding target image included in a video to which the still-image coding and the video coding are applied. A video coding unit (312) codes the coding target image by inter prediction coding that uses the decoded image as a reference image when the still-image coding is applied to the mage that is one image previous to the coding target image and the video coding is applied to the coding target image.