Intra Prediction Mode Signaling Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression standards, such as HEVC, face challenges in reducing the bit rate and computational complexity associated with signaling intra prediction modes, particularly when using a large number of prediction modes, which increases the cost of prediction mode signaling.
Innovation Solution
The method involves categorizing prediction modes and eliminating categories that are not adapted to the current block, updating the remaining categories by removing common modes, and signaling the selected prediction mode based on the updated categories, thereby optimizing prediction mode signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of intra prediction modes are used to predict a current block, then the prediction quality is improved, but the cost of signaling an intra prediction mode increases
Solution Approach 1:
The patent divides the 35 intra prediction modes into three categories (first category with 1 mode, second category with 19 modes, third category with 35 modes) based on the variance of reference pixels. This segmentation allows the system to adaptively select the appropriate category for each block, using only the necessary number of modes for that specific content type, thereby reducing the average bit rate for mode signaling while maintaining prediction quality when needed.
Solution Approach 2:
The patent dynamically adjusts the number of prediction modes used based on the variance of reference pixels for each current block. When variance is low, only the first category (1 mode) is used; when variance is medium, the second category (19 modes) is used; when variance is high, the third category (35 modes) is used. This dynamic adaptation optimizes the balance between prediction quality and signaling cost for each block individually.
2Measurement precision
If a large number of intra prediction modes are used to predict a current block, then the prediction quality is improved, but the complexity of selecting a prediction mode increases
Solution Approach 1:
By segmenting the prediction modes into three categories based on reference pixel variance, the patent reduces the search space for mode selection. Instead of evaluating all 35 modes for every block, the system only evaluates the modes appropriate for the detected variance level, significantly reducing computational complexity while maintaining the ability to achieve high prediction quality when necessary.
Solution Approach 2:
The dynamic categorization based on variance allows the system to adapt the complexity of mode selection to the actual content characteristics. For homogeneous regions (low variance), the selection process is simple (1 mode); for moderately complex regions (medium variance), 19 modes are evaluated; for highly complex regions (high variance), all 35 modes are considered. This dynamic approach optimizes the trade-off between prediction quality and selection complexity.
Data Source
AI summary
A method for decoding a coded data stream representative of at least one image, the image being divided into blocks, a set of at least two prediction modes being available for predicting a current block. The method includes: obtaining at least two categories of prediction modes from the set; eliminating at least one category in accordance with at least one first predetermined elimination criterion; updating the categories not eliminated by deleting prediction modes that are common with the at least one eliminated category; and decoding, from the data stream, an information identifying for the current block a prediction mode among the modes of the updated categories.