Reference Picture List Coding for High-Resolution Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression technologies face challenges in efficiently processing high-spatial resolution, high-frame rate video content due to increased memory storage and processing demands, particularly in constructing and managing reference picture lists for inter prediction.
Innovation Solution
A method and device for constructing a reference picture list by determining the type of reference pictures (short-term or long-term) based on picture order count differences and modulo values, optimizing the number of entries, and reducing syntax elements through efficient marking and coding of this information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the reference picture list includes more entries to improve prediction accuracy, then prediction precision is improved, but memory storage requirements and processing complexity increase
Solution Approach 1:
The reference picture list is segmented into two distinct types: short-term reference pictures (STRP) and long-term reference pictures (LTRP). This segmentation allows the system to manage reference pictures differently based on their temporal characteristics, reducing processing complexity while maintaining prediction accuracy through targeted management of each reference type.
Solution Approach 2:
The patent introduces specific parameters to distinguish and manage reference picture types: `st_ref_pic_flag` for identifying STRP entries and `abs_delta_poc_st` for storing absolute delta POC values. These parameter changes enable efficient encoding and decoding of reference picture information, reducing the bitstream complexity while maintaining the ability to access multiple reference pictures for accurate prediction.
2Measurement precision
If the reference picture list is constructed with more detailed information, then prediction accuracy is improved, but the amount of syntax elements and data to be processed increases
Solution Approach 1:
The patent extracts and separates critical reference picture information into specific syntax elements: `st_ref_pic_flag` for STRP identification and `abs_deltaPocSt` for temporal distance encoding. By extracting only the necessary information for STRP management and using inferred values for LTRP, the system reduces data complexity while preserving prediction accuracy.
Solution Approach 2:
Instead of explicitly coding all reference picture parameters, the patent uses inversion logic where LTRP entries are marked with a flag (`st_ref_pic_flag=0`) and their POC values are inferred from the current picture POC minus the stored absolute delta. This approach reduces the amount of explicit data needed while maintaining complete reference picture information for accurate prediction.
3Productivity
If the number of reference picture entries is increased to handle high-frame rate content, then prediction capability is improved, but memory storage requirements increase
Solution Approach 1:
The patent implements dynamic reference picture management where the DPB (decoded picture buffer) size can be adjusted based on the actual number of STRP and LTRP entries needed. The `num_strp_entries` and `num_ltrp_entries` parameters allow flexible allocation of memory resources, enabling the system to handle high-frame rate content with variable reference picture counts without requiring fixed, excessive memory capacity.
Data Source
AI summary
Embodiments of the disclosure provide methods and devices for decoding video signals using inter prediction. According to an embodiment of the disclosure, a method for processing a video signal comprises constructing a reference picture list of a current picture in the video signal and performing a prediction for a current picture by using the reference picture list, wherein constructing the reference picture list comprising, if a first entry of the reference picture list corresponds to a short-term reference picture (STRP), obtaining a picture order count (POC) difference between a picture related to the first entry and another picture, and if a second entry of the reference picture list corresponds to a long-term reference picture (LTRP), obtaining a POC modulo value of a picture related to the second entry. A reference picture list for identifying a picture may be generated in a simplified and effective manner.