Interpolation Filter for Non-Translational Video Motion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in achieving high video coding efficiency due to the complexity of interpolation for non-translational motion models, which results in reduced accuracy and increased computational complexity.
Innovation Solution
An improved interpolation filter is introduced for inter prediction in video coding, using a processing unit to determine motion vectors and apply spatial high-pass filters to sub-integer pixels and support pixels, allowing for precise interpolation and reduced memory bandwidth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional interpolation filters are used for non-translational motion models, then computational complexity is reduced, but prediction accuracy deteriorates
Solution Approach 1:
The patent divides the interpolation process into multiple stages: first generating intermediate half-integer pixel values, then using these as support for full-integer pixel interpolation. This segmentation allows the system to achieve high prediction accuracy through multi-pass filtering while keeping each individual filtering operation computationally simple, thus resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent performs preliminary interpolation to generate half-integer pixel values before using them as support pixels for subsequent full-integer pixel interpolation. This preliminary action creates a more accurate reference grid that improves final prediction accuracy while the filtering operations themselves remain computationally manageable through reuse of intermediate results.
2Measurement precision
If high-precision interpolation is applied to sub-integer pixels, then prediction accuracy is improved, but memory bandwidth requirements increase
Solution Approach 1:
The patent merges the interpolation of half-integer pixels with the subsequent full-integer pixel interpolation by using the half-integer values as support pixels. This combining eliminates the need to store separate high-precision intermediate results in memory, as they are immediately reused as filter support, thus reducing memory bandwidth requirements while maintaining high interpolation accuracy.
Solution Approach 2:
The interpolated half-integer pixel values serve dual purposes: they are both the target of interpolation and the support pixels for subsequent full-integer pixel interpolation. This self-service approach allows the system to achieve high precision without additional memory bandwidth, as the intermediate results automatically serve as references for the next processing stage.
3Productivity
If conventional filters are used for inter prediction, then device complexity is minimized, but video coding efficiency deteriorates
Solution Approach 1:
The patent implements a dynamic multi-pass filtering approach where the filter support is adaptively extended from integer pixels to include half-integer pixels. This dynamic extension allows the system to achieve superior video coding efficiency by capturing non-translational motion more accurately, while the filtering operations themselves remain structurally similar to conventional filters, thus limiting the increase in device complexity.
Data Source
AI summary
An apparatus for inter prediction of a sample value of a current full-integer pixel of a plurality of pixels associated with a current block of a current frame of a video signal is disclosed. The apparatus is configured to: determine a motion vector of the current full-integer pixel based on a motion compensation model; determine for the current full-integer pixel a corresponding sub-integer pixel in a reference frame; determine positions of a set of corresponding filter support pixels in the reference frame; obtain a respective sample value of the corresponding sub-integer pixel of the current full-integer pixel and the corresponding filter support pixels in the reference frame by applying a bilinear interpolation; and determine the inter predicted sample value of the current full-integer pixel.