Two-Layer Prediction for Intra-Frame Coding Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video codecs face inefficiencies in predicting prediction modes during intra-frame compression, especially with increased prediction modes, leading to higher overhead costs that offset bit savings, particularly when using 36 prediction modes.
Innovation Solution
A two-layer prediction method is introduced, where the first layer predicts a group-number based on neighboring group-numbers and the second layer predicts an index-number based on neighboring index-numbers, allowing for efficient encoding and decoding of prediction modes with reduced overhead by using flags to indicate matching or non-matching predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional one-layer prediction method is used to predict prediction modes, then the encoding process is simple, but the overhead bits increase when using 36 prediction modes, offsetting bit savings
Solution Approach 1:
The patent segments the prediction mode selection into two independent layers: first layer predicts group-number (dividing 36 modes into groups), second layer predicts index-number (selecting specific mode within group). This segmentation reduces overhead bits by enabling more efficient prediction and coding of the prediction mode information.
Solution Approach 2:
The patent introduces a two-dimensional prediction structure with group-number and index-number layers, transforming the single-layer prediction into a hierarchical two-layer approach. This dimensional change enables better compression by separating the prediction space into coarse (group) and fine (index) levels.
2Measurement precision
If 36 prediction modes are used to improve coding precision, then prediction accuracy improves, but overhead bits increase offsetting the bit savings
Solution Approach 1:
By segmenting the 36 prediction modes into groups (group-number prediction) and then selecting specific modes within groups (index-number prediction), the patent reduces the total overhead bits required to signal prediction modes while maintaining access to all 36 modes for high prediction accuracy.
Solution Approach 2:
The two-layer hierarchical structure creates a dimensional breakdown of prediction mode selection, where the first dimension (group-number) captures coarse prediction information and the second dimension (index-number) refines the selection, reducing overall bit cost while preserving prediction accuracy.
3Productivity
If independent two-layer prediction is used to reduce overhead bits, then coding efficiency improves, but the encoding and decoding process becomes more complex
Solution Approach 1:
The patent segments the prediction process into two independent layers that can be processed separately and in parallel, reducing overhead bits while managing complexity through modular design where each layer handles a specific aspect of prediction mode selection.
Solution Approach 2:
By introducing the two-dimensional hierarchical prediction structure, the patent achieves better coding efficiency through reduced overhead bits, and manages the increased processing complexity through clear separation of concerns between group-number and index-number prediction layers.
Data Source
AI summary
An apparatus comprising a processor configured to receive a current block of an image, select an optimal prediction mode of a plurality of prediction modes for the current block, wherein the selected prediction mode corresponds to a selected one of a plurality of group-numbers and a selected one of a plurality of index-numbers, predict one of the plurality of group-numbers based on a first known group-number and a second known group-number, wherein the predicted group-number is predicted independent from any known index-numbers, clear a first flag when the predicted group-number matches the selected group-number, and set the first flag and encode the selected group-number into a coded data stream when the predicted group-number does not match the selected group-number, wherein the first flag is located in an overhead of the coded data stream.