Adaptive Interpolation for Image Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image prediction methods, such as Motion Estimation (ME) and Motion Compensation (MC), are limited in accurately predicting future frames, especially for non-stationary processes and non-Gaussian signals, as they assume translation motion and do not fully exploit all available information from past frames.
Innovation Solution
A method that computes predicted frames by defining collocated blocks in reference frames, deriving motion vectors, and calculating adaptive interpolation coefficients to transform pixels from one frame to another, allowing for accurate prediction of pixels in the predicted frame using coefficients derived from high similarities between adjacent frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Motion Estimation and Motion Compensation are used for image prediction, then the prediction process can be implemented with simple translation motion assumption, but the prediction accuracy deteriorates for non-stationary processes and non-Gaussian signals
Solution Approach 1:
The patent applies dynamics by transitioning from static translation-based motion compensation to dynamic adaptive filtering. The filter coefficients are adaptively adjusted based on local signal characteristics (variance, correlation) to match the actual motion and statistical properties of each block, enabling the system to handle non-stationary and non-Gaussian signals effectively while maintaining implementation feasibility.
Solution Approach 2:
The patent changes parameters by deriving adaptive filter coefficients from local signal statistics rather than using fixed translation parameters. The system calculates variance and correlation parameters for each block and uses these to determine optimal filtering parameters, thereby improving prediction accuracy for diverse signal types without significantly increasing system complexity.
2Loss of information
If Motion Estimation and Motion Compensation are used, then fewer bits are needed to describe the residual, but the method cannot fully exploit all available information from past frames for non-Gaussian processes
Solution Approach 1:
The patent changes the approach by using adaptive filter coefficients derived from local signal statistics instead of fixed motion compensation parameters. This allows the system to fully exploit local correlations and redundancies in non-Gaussian processes, improving prediction reliability while maintaining efficient bit rate reduction through targeted adaptation to each block's characteristics.
Solution Approach 2:
The patent applies local quality by computing separate adaptive filter coefficients for each block based on its local signal characteristics. Each block is processed with filtering parameters optimized for its specific variance and correlation properties, enabling the system to exploit all available local information while maintaining overall compression efficiency.
3Measurement precision
If adaptive interpolation coefficients are computed from local signal characteristics, then prediction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies local quality by computing adaptive filter coefficients for each block based on its local signal characteristics (variance, correlation). This localized approach improves prediction accuracy for each block while limiting the overall computational complexity increase, as the complex adaptive computation is performed only where needed rather than globally.
Solution Approach 2:
The patent applies partial action by computing adaptive coefficients selectively for blocks that benefit most from it, rather than uniformly applying complex adaptive filtering to all blocks. This allows the system to achieve improved prediction accuracy for critical blocks while controlling overall computational complexity through selective application.
Data Source
AI summary
A method is provided for computing a predicted frame from a first and a second reference frames, the method comprising, for each block of pixels in the predicted frame, the acts of defining a first block of pixels in the first reference frame collocated with a third block of pixels which is the block of pixels in the predicted frame; defining a second block of pixels corresponding, in the second reference frame, to the first block of pixels along the motion vector of the first block from the first to second reference frames; computing a first set of coefficients allowing the transformation of the pixels of the first block into pixels of the second block; computing pixels of the third block using the first set of coefficients and pixels from a fourth block collocated in the first reference frame with the second block of pixels.