Wide-Angle Intra Prediction for Non-Square Video Blocks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image and video coding technologies lack effective wide-angle intra prediction methods, particularly for non-square blocks, leading to suboptimal reference sample utilization and reduced coding efficiency.

Innovation Solution

Introduce wide-angle intra prediction modes that extend beyond traditional 45-degree angles, utilizing reference samples at obtuse angles relative to the top-left direction, and encode these modes directly in the bitstream, with the choice of primary and wide-angle modes determined by the block shape.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional intra prediction modes (PModes) of 65 modes are used, then the coding complexity is manageable, but the prediction accuracy for certain directional regions is insufficient leading to higher bitrate requirements

Engineering Contradiction:
Improveprediction accuracyVSAvoidbitrate efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The prediction mode space is segmented into two distinct sets: conventional PModes (65 modes) and additional wide-angle prediction modes (AWPModes). This segmentation allows the system to use different prediction strategies for different directional requirements, improving overall prediction accuracy without uniformly increasing complexity across all prediction scenarios.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The prediction mode selection is made dynamically based on the actual prediction direction required. The encoder determines whether to use a conventional PMode or an AWPMode depending on the angular position and characteristics of the block being predicted, allowing adaptive optimization of prediction accuracy for each specific case.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If more prediction modes are added to improve coverage of all directions, then the prediction accuracy improves, but the coding complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcoding complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Different prediction mode sets are applied locally based on directional requirements. Conventional PModes are used for directions where they provide sufficient accuracy, while AWPModes are selectively applied for wide-angle directions where conventional modes fail. This local differentiation maintains coding simplicity for most cases while providing enhanced accuracy where needed.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The invention changes the parameter of prediction mode selection by introducing a dual-mode system where the mode type (conventional vs. wide-angle) is selected based on directional parameters. This parameter-based selection allows the system to adapt to different prediction scenarios without requiring a complete redesign of the prediction framework.

Inventive Principle:
Principle #35Parameter changes

3Speed

If conventional intra prediction is used for all blocks, then the processing speed is maintained, but the prediction quality for wide-angle directions deteriorates

Engineering Contradiction:
Improveprocessing speedVSAvoidprediction quality
Core Design Contradiction:
SpeedVSMeasurement precision

Solution Approach 1:

The prediction system dynamically selects between conventional and wide-angle modes based on the actual prediction direction. This dynamic selection ensures that processing speed is maintained for the majority of cases using conventional modes, while prediction quality is improved for wide-angle directions by switching to appropriate AWPModes when needed.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4383704B1Video and image coding with wide-angle intra prediction
Publication Date: 2026.05.20 NOKIA TECHNOLOGIES OY
  • EP4383704B1 patent drawingFigure 1a~1b
  • EP4383704B1 patent drawingFigure 2a
  • EP4383704B1 patent drawingFigure 2b

AI summary

The invention relates to a method for decoding, comprising receiving a bitstream comprising encoded image data, forming from the bitstream in a computer memory a set of reference pixels having reference pixel values for intra-picture directional prediction, predicting a pixel value in a predicted image block by intra-picture directional prediction by using values of one or more selected reference pixels in forming a predicted a pixel value, wherein the selected reference pixels are selected from a prediction direction with respect to the predicted pixel, and the prediction direction being a wide-angle prediction direction that forms an obtuse angle with top-left direction of the predicted pixel, repeating the pixel prediction for a number of pixels to form a predicted image block, and using the predicted image block in decoding an image block from the bitstream to obtain a decoded image block. A corresponding method for decoding as well as a decoder, an encoder, computer program products for the same purpose, and a signal are provided.