Intra-Prediction Geometry Transform Inference for Lower Video Signaling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The existing video compression systems, such as HEVC and VVC, incur additional signaling cost due to the use of Reconstruction-Reordered IBC (RR-IBC) mode for screen content coding, which reduces coding efficiency.

Innovation Solution

Implement a method for intra-prediction that determines the geometry transform type for a block based on the intra-prediction mode of neighboring blocks, inferring the transform type from the neighboring block's geometry transform, and signal the best geometry transform type in the bitstream.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If RR-IBC mode is used for screen content coding, then coding capability for screen content is improved, but signaling overhead increases

Engineering Contradiction:

Improvecoding efficiencyVSAvoidsignaling overhead

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The geometry transform type for the current block is determined self-service by checking the intra-prediction modes of neighboring blocks and inferring the transform type from them, eliminating the need for separate signaling of the geometry transform type parameter

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The geometry transform type is copied from neighboring blocks that use the same intra-prediction mode. The decoder copies the transform type information from already-decoded neighboring blocks to the current block, avoiding redundant signaling

Inventive Principle:

Principle #26Copying

2Manufacturing precision

If geometry transform type is signaled for each block, then prediction accuracy is improved, but bitstream overhead increases

Engineering Contradiction:

Improveprediction accuracyVSAvoidbitstream size

Core Design Contradiction:

Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The geometry transform type determination is merged with the existing intra-prediction mode decoding process. By combining these two functions, the patent avoids separate signaling of transform type and maintains prediction accuracy through context-aware transform selection

Inventive Principle:

Principle #5Merging (Combining)

3Adaptability or versatility

If multiple geometry transform types are supported, then adaptability to different content is improved, but decoding complexity increases

Engineering Contradiction:

Improvecontent adaptabilityVSAvoiddecoding complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The geometry transform type is dynamically selected based on the intra-prediction mode of the current block and its neighboring blocks. Different transform types (identity, horizontal flip, vertical flip, transpose) are applied dynamically according to the prediction context, maintaining adaptability while simplifying the selection process

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20260101044A1Encoding/decoding video picture data

Publication Date: 2026.04.09 BEIJING XIAOMI MOBILE SOFTWARE CO LTD

AI summary

The present invention relates to a method of intra-predicting a block of a video picture according to a first intra-prediction mode using a first geometry transform identified by a first geometry transform type. The method determines whether at least one neighboring block is predicted according to a second intra-prediction mode using a second geometry transform identified by a second geometry transform type and, if at least one neighboring block is predicted according to the second intra-prediction mode, the first geometry transform type is inferred from said second geometry transform type.