Video Encoder POC Signaling for Reference Picture Ambiguity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The HEVC standard faces ambiguity in determining which long-term reference pictures to keep in the decoded picture buffer due to identical least significant bits of Picture Order Count (POC) values, leading to unclear reference picture usage and error resilience issues during video decoding.
Innovation Solution
The encoder and decoder implement a method to unambiguously identify reference pictures by ensuring the full POC is signaled when multiple pictures share the same least significant bits of POC, using a parameter to indicate whether the full POC or least significant bits should be used, and check for potential errors in the video stream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If only the least significant bits (lsb) of POC are signaled in the RPS, then the bitstream size is reduced and decoding is simplified, but ambiguity arises when multiple long-term reference pictures have identical POC lsb values, making it unclear which pictures to keep in the DPB
Solution Approach 1:
The patent applies dynamics by making the POC signaling method adaptive rather than fixed. The encoder dynamically selects between signaling full POC or only POC lsb based on the actual content of the reference picture list. When multiple pictures share the same POC lsb, full POC is signaled to ensure uniqueness; otherwise, only POC lsb is signaled to save bandwidth. This dynamic adaptation resolves the contradiction between compression efficiency and identification accuracy.
Solution Approach 2:
The patent changes the parameter being signaled in the RPS from a fixed choice (either always full POC or always POC lsb) to a variable choice. The parameter signaled depends on the specific conditions of the reference picture list content. This parameter change allows the system to optimize between bitstream size and reference picture identification accuracy on a per-picture basis, resolving the technical contradiction.
2Reliability
If the full POC is always signaled in the RPS, then reference picture identification is unambiguous even when multiple pictures share the same POC lsb, but the bitstream size increases and decoding complexity increases
Solution Approach 1:
The patent makes the decoding process dynamic by conditionally processing different POC formats. The decoder checks whether multiple reference pictures share the same POC lsb value and adapts its processing accordingly. When such ambiguity is detected, it uses full POC values for identification; otherwise, it processes only POC lsb values. This dynamic approach reduces decoding complexity while maintaining reliable reference picture identification.
Solution Approach 2:
The patent applies self-service by enabling the decoding system to automatically detect and handle the ambiguity condition without external intervention. The decoder autonomously determines whether full POC or POC lsb should be used based on the reference picture list content, eliminating the need for complex pre-configuration or external control mechanisms while maintaining identification accuracy.
3Adaptability or versatility
If multiple reference pictures with identical POC lsb values are allowed in the DPB, then more reference pictures can be utilized for inter prediction, but ambiguity arises in determining which pictures to keep when RPS indicates a picture by POC lsb
Solution Approach 1:
The patent resolves the ambiguity by adding another dimension to the POC representation. When multiple pictures share the same POC lsb, the full POC value (which includes higher-order bits) is used as an additional distinguishing dimension. This dimensional expansion allows the system to maintain multiple reference pictures with identical POC lsb values while still being able to unambiguously identify and manage them through the full POC value when needed.
Solution Approach 2:
The patent effectively creates a composite identification mechanism that combines both POC lsb and full POC values. The system uses POC lsb for normal operation and compression, but combines it with full POC values when ambiguity is detected, creating a composite identification approach that leverages the advantages of both representations to maintain both versatility and clarity.
Data Source
AI summary
According to an embodiment, an encoder is configured to ensure that for each long-term picture of a RPS of a picture i the value of a flag referred to as delta_poc_msb_present_flag[i] is equal to 1 when there are at least two reference pictures in a decoded picture buffer with lsb's of the POC referred to as pic_order_cnt_lsb equal to the lsb's of the respective long-term picture i denoted POC lsbLt[i]. When the delta_poc_msb_present_flag is equal to 1, the long-term picture is indicated by the full POC.