Motion Vector-Based Filter Selection for Video Motion Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face reduced prediction accuracy and increased data rates due to varying motion blur, which complicates motion compensation and increases the number of signaling bits required for blur variation between frames.
Innovation Solution
A video coder and decoder system that utilizes motion vectors to select and apply filters to reference frames based on the orientation of motion vectors, eliminating the need for additional signaling bits by integrating filter information into the motion information within the bit stream, and allowing for adaptive filtering at various granularities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If motion compensation is applied to predict frames in video coding, then compression efficiency is improved, but prediction accuracy deteriorates due to varying motion blur between frames
Solution Approach 1:
The patent applies dynamic filtering to the reference frame based on motion vector characteristics. Instead of using a static reference frame, the system dynamically adjusts the reference frame by applying filters (such as Gaussian filters or motion blur filters) whose parameters are determined by the motion vector's magnitude and orientation. This dynamic adjustment allows the reference frame to adapt to varying motion conditions, thereby maintaining prediction accuracy while preserving compression efficiency.
Solution Approach 2:
The patent changes parameters of the reference frame based on motion vector analysis. Specifically, it modifies the reference frame's characteristics (such as applying blur filters with specific kernel sizes and sigma values) according to the motion vector's properties. This parameter change approach allows the system to compensate for motion blur effects by adjusting the reference frame to match the expected motion characteristics, thus improving prediction accuracy without sacrificing compression performance.
2Measurement precision
If additional signaling bits are used to indicate blur variation between frames, then prediction accuracy is improved, but data rate increases
Solution Approach 1:
The patent employs self-service by using existing motion vector information (which is already present in the bit stream for inter-prediction) to derive filter parameters for reference frame filtering. Instead of requiring additional signaling bits to communicate blur variation information, the system extracts the necessary information from the motion vectors themselves. This approach allows the system to improve prediction accuracy through adaptive filtering while avoiding any increase in data rate, as no additional bits are required.
Solution Approach 2:
The patent makes the motion vector serve multiple functions: it not only provides motion compensation information for prediction but also serves as the basis for determining filter parameters for reference frame filtering. This multi-functionality approach eliminates the need for separate signaling mechanisms for blur variation, as the motion vector already conveys the necessary information. Consequently, the system achieves improved prediction accuracy without increasing the data rate, since the same motion vector data is reused for dual purposes.
3Measurement precision
If filtering is applied to reference frames to compensate for motion blur, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies filtering locally and selectively to reference frames based on motion vector characteristics. Instead of applying complex filtering operations to entire frames uniformly, the system applies filters only to regions where motion blur is significant, and chooses filter parameters (such as kernel size and sigma) based on local motion characteristics. This localized approach reduces the overall computational complexity while maintaining prediction accuracy in the critical regions affected by motion blur.
4Quantity of substance
If motion vectors are used for filter selection, then signaling bits are reduced, but filter selection accuracy may deteriorate
Solution Approach 1:
The patent uses motion vector information partially or selectively for filter selection rather than relying on it completely. The system analyzes the motion vector's properties (magnitude and orientation) and uses this information to guide filter parameter selection, but may also incorporate additional considerations or use default filters when motion vector information is insufficient or ambiguous. This partial use of motion vectors maintains filter selection accuracy while avoiding the need for additional signaling bits, as the system does not require perfect motion vector information to make effective filter choices.
Data Source
AI summary
A frame buffer is adapted to store at least one reference frame of a video stream, said reference frame being different from a current frame of the video stream. A motion estimation unit is adapted to generate a motion vector by estimating motion between a reference block of the reference frame and a current block of the current frame. A blurring filter selection unit is adapted to select one of a plurality of blurring filters depending on the orientation of the motion vector. A filtering unit is adapted to generate a filtered reference block by applying the selected blurring filter to the reference block. A motion compensation unit is adapted to generate a filtered prediction block of the current block on the basis of the current block, the motion vector and the filtered reference block.