Geometric-Transformation Motion Compensation Prediction Unit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image coding methods using motion compensation prediction, such as those in the MPEG series, face inefficiencies in compressing coding amounts, particularly when geometric transformation is employed, as they do not effectively manage motion vector information to minimize coding amounts and distortion.
Innovation Solution
An image coding apparatus and method that calculates motion vectors for representative pixels in a target block and interpolates for other pixels, using geometric-transformation motion compensation prediction to select optimal prediction modes, and codes the difference motion vectors and prediction error signals to reduce overall coding amounts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If motion compensation prediction by geometric transformation is used, then coding amount compression is improved, but device complexity increases due to multiple prediction modes
Solution Approach 1:
The patent divides the block motion compensation process into multiple prediction modes (first mode with one motion vector, second mode with two motion vectors, third mode with three motion vectors, and fourth mode with four motion vectors). Each mode segments the motion representation differently, allowing the system to select the most efficient mode for each block, thereby reducing overall coding amount while managing complexity through structured segmentation.
Solution Approach 2:
The patent dynamically selects among multiple prediction modes based on the characteristics of each block. The prediction mode determination unit chooses the optimal mode (first through fourth modes) for each block, allowing the system to adapt to different motion patterns and geometric transformations, improving coding efficiency without requiring all modes to be used uniformly.
2Measurement precision
If multiple prediction modes are used for geometric transformation, then prediction accuracy is improved, but calculation complexity increases
Solution Approach 1:
The patent segments the motion prediction into four distinct modes with increasing numbers of motion vectors (one, two, three, or four). This segmentation allows the system to achieve higher prediction accuracy for complex geometric transformations by using modes with more vectors when needed, while maintaining lower calculation complexity for simpler cases by using modes with fewer vectors.
Solution Approach 2:
The patent changes the parameter of motion vector quantity across different prediction modes. By varying the number of motion vectors from one to four depending on the block characteristics and transformation complexity, the system optimizes the balance between prediction accuracy and calculation complexity, using more vectors only when the transformation geometry requires it.
3Measurement precision
If motion vectors for all pixels are calculated directly, then prediction precision is improved, but processing time increases
Solution Approach 1:
The patent segments the set of pixels into representative pixels (at block vertices) and other pixels. Motion vectors are calculated directly only for representative pixels, while motion vectors for other pixels are derived through interpolation. This segmentation significantly reduces the number of direct calculations required while maintaining prediction precision through the interpolation process.
Solution Approach 2:
The patent uses interpolation as an intermediary process to derive motion vectors for non-representative pixels from the motion vectors of representative pixels. This intermediary approach avoids the need for direct motion vector calculation for every pixel, reducing processing time while maintaining adequate precision through the mathematical interpolation of motion information.
4Productivity
If representative pixels are selected for motion vector calculation, then processing efficiency is improved, but prediction accuracy may deteriorate
Solution Approach 1:
The patent applies local quality by treating representative pixels (at block vertices) with direct motion vector calculation while using interpolation for other pixels. This local differentiation in processing quality maintains high accuracy at critical block boundaries while improving overall processing efficiency, recognizing that vertex regions require more precise treatment than interior regions.
Solution Approach 2:
The patent uses interpolation as an intermediary to extend the motion information from representative pixels to the entire block. This intermediary process preserves prediction accuracy by mathematically deriving motion vectors for non-representative pixels based on the precisely calculated vectors at representative locations, maintaining fidelity without requiring direct calculation for every pixel.
Data Source
AI summary
A geometric-transformation motion compensation prediction unit calculates, for each of a plurality of prediction modes, a motion vector and a prediction signal between a target block in a target image and a reference block in a reference image obtained by performing geometric transformation on the target block, selects pixels located at vertices constituting the target block, pixels located near the vertices, or interpolation pixels located near the vertices as representative pixels corresponding to the vertices in each prediction mode, calculates the respective motion vectors of these representative pixels, and calculates the respective motion vectors of pixels other than the representative pixels by interpolation using the motion vectors of the representative pixels so as to calculate the prediction signal.