Intra Prediction Mode Selection Using Reference Region Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image encoding and decoding methods are inefficient as they do not effectively consider the characteristics of reference regions during intra prediction, leading to suboptimal selection of prediction modes and inefficient encoding processes.
Innovation Solution
A method and apparatus that determine prediction units and reference regions within an image frame, analyze sample values in the reference regions, and adjust sample values based on these analyses to improve intra prediction during encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If all prediction modes are expressed in the same probability using fixed length coding, then the encoding process is simple, but the encoding and decoding efficiency deteriorates because it does not distinguish between frequently and infrequently used prediction modes
Solution Approach 1:
The patent changes the encoding parameter from fixed length coding to variable length coding based on probability. Frequently used prediction modes are assigned shorter codes while infrequently used modes get longer codes, optimizing the balance between encoding simplicity and efficiency.
Solution Approach 2:
The patent introduces a probability estimation mechanism that analyzes the frequency of prediction mode usage and adjusts the coding strategy accordingly. This feedback loop enables the system to adaptively optimize encoding efficiency based on actual usage patterns.
2Ease of operation
If intra prediction is performed using the same method for prediction units with the same modes, then the decoding process is simplified, but the prediction accuracy deteriorates because it does not consider the characteristics of reference regions
Solution Approach 1:
The patent applies different prediction methods to different prediction units based on their reference region characteristics. Instead of using a uniform approach, the system analyzes local properties such as gradient and variance of reference regions to select the most appropriate prediction method for each specific unit, thereby improving prediction accuracy while maintaining reasonable decoding complexity.
3Device complexity
If conventional intra prediction methods are used without considering reference region characteristics, then the encoding and decoding process remains simple, but the overall performance deteriorates
Solution Approach 1:
The patent performs preliminary analysis of reference region characteristics (such as gradient and variance calculations) before selecting the prediction method. This advance preparation enables the system to make informed decisions about which prediction technique to use, improving overall performance without significantly increasing the complexity of the main encoding and decoding processes.
Data Source
AI summary
Provided is a method of decoding an image, the method including: determining at least one prediction unit included in a current frame that is one of at least one frame forming the image; determining a reference region to be referred to by a current prediction unit that is one of the at least one prediction unit; changing a sample value included in at least one of the current prediction unit and the reference region, based on an analyzing result of a sample value of the reference region; determining a sample value included in the current prediction unit, based on a result of changing the sample value; and decoding the image based on the determined sample value of the current prediction unit.