Reference Picture Signaling for Random Access in Image Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for decoding random-accessed pictures in image sequences require decoding all reference pictures starting from the previous IDR picture, leading to inefficiencies and increased file size due to poor coding efficiency when images are represented as series of still pictures.
Innovation Solution
An improved method for identifying and signaling reference pictures by assigning unique identifiers to reference samples, determining unique combinations of sample identifiers and reference sample identifiers, and mapping these to indices, allowing only necessary pictures to be decoded for random access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If all reference pictures starting from the previous IDR picture are decoded to enable random access, then random access capability is achieved, but decoding delay and file size increase significantly
Solution Approach 1:
The patent segments the reference picture set into multiple subsets, each associated with specific prediction modes. Instead of decoding all reference pictures, the decoder only decodes the subset corresponding to the accessed picture's prediction mode, thereby reducing decoding delay while maintaining random access capability
Solution Approach 2:
The patent applies partial action by decoding only the necessary portion of reference pictures (specific subset) rather than all reference pictures. This partial decoding approach achieves the required random access functionality without the excessive overhead of decoding the entire reference picture set
2Reliability
If all reference pictures are decoded to ensure correct decoding of any accessed picture, then decoding accuracy is maintained, but file size and processing complexity increase
Solution Approach 1:
The reference picture set is divided into multiple subsets based on prediction mode requirements. The encoder signals which subset should be decoded, allowing the decoder to process only the relevant subset and maintain decoding accuracy without handling all reference pictures, thus reducing processing complexity
Solution Approach 2:
The encoder performs preliminary action by determining and signaling the appropriate reference picture subset before decoding. This allows the decoder to prepare and decode only the necessary reference pictures in advance, ensuring decoding accuracy while minimizing processing complexity
3Reliability
If all reference pictures are decoded to support random access, then complete picture reconstruction is ensured, but coding efficiency deteriorates
Solution Approach 1:
The patent segments the reference picture processing into multiple subsets, allowing the decoder to reconstruct pictures correctly by decoding only the relevant subset rather than all reference pictures. This segmentation maintains reconstruction completeness while improving coding efficiency by avoiding redundant decoding operations
4Reliability
If the subset of required reference pictures cannot be easily identified, then random access becomes inefficient, but decoding all pictures ensures correctness
Solution Approach 1:
The encoder performs preliminary action by determining and signaling which reference picture subset is required for each picture before decoding. This preliminary identification eliminates the need for complex runtime analysis and enables fast random access while ensuring decoding correctness through proper subset selection
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method for signaling reference pictures of a picture sequence encoded in a bitstream, the method comprising classifying pictures of the picture sequence into reference pictures and non-reference pictures; providing each reference picture with a unique picture identifier; determining, at least for inter-coded picturesof the picture sequence, all unique combinations of the picture identifier and at least one reference picture identifier, where the at least one reference picture identifier for a particular picture identifies all pictures that may be used as a reference for predicting said particular picture;and assigning intra-coded reference pictures at least with their unique picture identifier and the inter-coded pictures with their corresponding unique combination of the picture identifier and the at least one reference picture identifier