Reference Frame Screening Model for HEVC Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current reference frame selection strategies in HEVC (High Efficiency Video Coding) are inefficient, as they rely on a sequential backward search from the reference frame chain list, which may not effectively identify the most relevant reference frames for video coding and decoding.
Innovation Solution
A method and apparatus for training a reference frame screening model using machine learning, where a training sample set comprising video frames and labels indicating reference frames is used to train the model. This model can then screen reference frames for a target video sequence, improving the selection process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sequential backward search from reference frame chain list is used, then reference frame selection can be performed, but selection efficiency is low and accuracy is insufficient
Solution Approach 1:
The patent replaces the traditional sequential backward search mechanical process with a machine learning model that automatically screens and selects reference frames. The model takes video sequences as input and outputs optimized reference frame selections, substituting the manual iterative search process with an intelligent system that achieves both higher accuracy and efficiency simultaneously.
Solution Approach 2:
The patent changes the selection criteria parameters by training a machine learning model on video content characteristics, motion patterns, and temporal relationships. Instead of fixed sequential selection, the model dynamically adjusts reference frame selection based on learned parameters from training data, enabling adaptive optimization of both accuracy and efficiency.
2Reliability
If all possible reference frames are stored in reference picture set, then complete reference information is available, but memory usage and processing complexity increase
Solution Approach 1:
The patent extracts only the most relevant reference frames from the complete reference picture set using the machine learning model. Instead of managing all possible reference frames, the system identifies and retains only those frames that are most useful for current video coding tasks, reducing memory usage and processing complexity while maintaining reliability through intelligent selection.
Solution Approach 2:
The patent applies partial action by selecting a subset of reference frames that are sufficient for achieving good coding performance without needing all available frames. The machine learning model determines the optimal number and selection of reference frames, avoiding the excessive action of processing and storing all possible reference frames while maintaining coding reliability.
Data Source
AI summary
The present disclosure provides a method and apparatus for training a reference frame screening model, a method and apparatus for screening a reference frame, an electronic device, a storage medium, and a computer program product, relates to the field of artificial intelligence. An implementation scheme is: acquiring a training sample set, where training samples in the training sample set include a sequence of video frames and labels corresponding to video frames in the sequence of video frames, and the labels are used to represent whether the video frames corresponding to the labels are reference frames of other video frames in the sequence of video frames; and training, using a machine learning method, using the sequence of video frames as input, using the labels corresponding to the input sequence of video frames as desired output, to obtain a reference frame screening model.