Video Coding Filter Selection for 360-Degree Image Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding methods for 360-degree images require significant storage and computational resources due to non-linear intra-prediction processes, which include feature detection, matrix-vector multiplication, and inverse discrete cosine transform operations, leading to high memory usage and complexity.
Innovation Solution
An image processing apparatus and method that selects adaptive filters based on the size of the coding unit, performs filtering operations on reference boundaries to generate filtering reference values, and uses interpolation to generate prediction values, reducing the need for complex operations like IDCT and matrix-vector multiplication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If non-linear intra-prediction processes (feature detection, matrix-vector multiplication, inverse discrete cosine transform) are used for 360-degree images, then prediction accuracy is improved, but memory storage and computational complexity increase significantly
Solution Approach 1:
The patent extracts and removes the complex non-linear intra-prediction processes (feature detection, matrix-vector multiplication, inverse discrete cosine transform) from the video coding system for 360-degree images. By eliminating these computationally intensive operations while maintaining the essential prediction functionality through alternative methods, the patent significantly reduces computational complexity and memory storage requirements while preserving adequate prediction accuracy for immersive video applications
Solution Approach 2:
The patent changes the parameters of the prediction process by transitioning from non-linear operations (IDCT, matrix-vector multiplication) to simpler linear operations. This parameter change involves modifying the mathematical operations used in intra-prediction, replacing complex transformations with more efficient calculations that achieve comparable prediction results with reduced computational burden and lower memory requirements
2Measurement precision
If non-linear intra-prediction processes are used for 360-degree images, then prediction accuracy is improved, but memory storage requirements increase
Solution Approach 1:
The patent extracts and removes the memory-intensive non-linear intra-prediction processes from the video coding system. By eliminating the need for storing intermediate results of feature detection, matrix-vector multiplication, and inverse discrete cosine transform operations, the patent significantly reduces memory storage requirements while maintaining the essential prediction functionality through more efficient methods
3Device complexity
If filter selection based on coding unit size is performed, then computational complexity is reduced, but filtering operation overhead is added
Solution Approach 1:
The patent applies local quality by selecting different filtering operations based on the specific size of each coding unit. Instead of applying a uniform filtering approach to all coding units, the patent adapts the filtering strategy to the local characteristics of each coding unit size, optimizing the balance between computational complexity and coding efficiency for different regions of the video data
Data Source
AI summary
A method and an image processing apparatus for video coding are provided. The method is adapted to the image processing apparatus and includes following steps. A current coding unit is received, and filter selection is performed according to a size of the current coding unit. At least one selected filter is used to perform a filtering operation on a plurality of reference boundaries of the current coding unit to generate a plurality of filtering reference values. An interpolation operation is performed on the current coding unit according to the filtering reference values to generate a plurality of interpolated prediction values.