Signaling Long Term Reference Pictures in Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques face inefficiencies in signaling long term reference pictures (LTRPs), particularly in the semantics of flag delta_poc_msb_present_flag and the ordering of LTRP signaling, which can increase bit usage and introduce bitstream conformance and interoperability issues.
Innovation Solution
The proposed techniques involve signaling LTRP identifying information in a slice header or sequence parameter set, ensuring non-decreasing or non-increasing order of least significant bits (LSBs) for POC values, and modifying the semantics of delta_poc_msb_present_flag to only depend on marked 'used for reference' pictures, reducing unnecessary MSB signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If LTRP signaling is performed with flexible ordering of POC values, then adaptability of reference picture selection is improved, but bitstream conformance and interoperability issues arise due to ambiguous semantics of delta_poc_msb_present_flag
Solution Approach 1:
The patent changes the parameter constraints by requiring LTRP POC values to be in non-decreasing order, transforming a flexible but ambiguous system into a constrained but unambiguous one. This ordering constraint resolves the semantic ambiguity of delta_poc_msb_present_flag while maintaining the adaptability of reference picture selection through the syntax element long_term_idx_sps that allows selective inheritance of LTRPs from SPS.
2Loss of information
If MSB signaling is performed for all LTRPs regardless of reference usage, then completeness of reference information is improved, but bit usage increases due to unnecessary MSB signaling
Solution Approach 1:
The patent extracts and separates the signaling of LTRP identification from full POC value signaling. By using the syntax element long_term_idx_sps to reference LTRPs in the SPS and only signaling LSBs when needed, the patent removes unnecessary MSB signaling for LTRPs that are already identified in the SPS, thereby reducing bit usage while maintaining reference information completeness.
3Measurement precision
If LTRP identification is performed using full POC values, then precision of reference picture identification is improved, but coding efficiency decreases due to increased bit usage
Solution Approach 1:
The patent segments the POC value into MSB and LSB components, signaling only the necessary LSB portion for LTRP identification. By dividing the POC value and selectively signaling only the least significant bits through the poc_lsb_lt syntax element, the patent maintains precise reference picture identification while significantly reducing the number of bits required, thereby improving coding efficiency.
Data Source
AI summary
A video coder codes a slice header for a slice of video data. The slice header includes a syntax element comprising identifying information for a long term reference picture, wherein the identifying information is explicitly signaled in the slice header or derived from a sequence parameter set corresponding to the slice. When the syntax element indicates that the identifying information for the long term reference picture is explicitly signaled, to code the slice header, the video coder is further configured to code a value for the identifying information for the long term reference picture in the slice header.