Video Error Concealment Using Diagonal Neighbor Pixels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing error concealment techniques in video codecs, such as H.264/AVC, often fail to properly handle corrupted video blocks due to the unavailability of direct neighbor macroblocks, leading to incomplete concealment and propagation of errors in video playback.
Innovation Solution
The proposed solution involves methods for concealing corrupted video blocks by utilizing available neighbor information without requiring direct neighbor subdivisions, including copying boundary pixels from available macroblocks and using motion vector recovery techniques like AVGMV and BESTMV to reconstruct corrupted inter-coded blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing error concealment techniques are used, then some corrupted blocks can be concealed, but direct neighbor macroblocks are unavailable leading to incomplete concealment and error propagation
Solution Approach 1:
The patent uses diagonal neighbor macroblocks as intermediary references instead of direct neighbors. By copying boundary pixels from diagonal neighbors (top-left, top-right, bottom-left, bottom-right macroblocks) and using them to reconstruct corrupted blocks, the system overcomes the unavailability of direct neighbor information while maintaining effective error concealment.
2Reliability
If motion vector recovery techniques are used, then inter-coded blocks can be reconstructed, but complexity increases due to multiple recovery methods (AVGMV, BESTMV)
Solution Approach 1:
The patent implements multiple motion vector recovery methods (AVGMV and BESTMV) but applies them selectively based on availability of reference blocks. The system attempts BESTMV first (more accurate but computationally intensive), and falls back to AVGMV (simpler) when reference blocks are unavailable, thus achieving reliable reconstruction without always incurring maximum complexity.
3Reliability
If boundary pixels are copied from diagonal neighbors, then concealed blocks improve, but processing time increases due to additional pixel copying operations
Solution Approach 1:
The patent performs boundary pixel copying from diagonal neighbor macroblocks as a preliminary step before main reconstruction operations. By preparing the reference pixels in advance from available diagonal neighbors, the system enables faster subsequent reconstruction operations and avoids repeated access to the same reference data during error concealment processing.
Data Source
AI summary
Corrupted subdivisions of streaming data are concealed using a matching subdivision in a previously reconstructed reference picture. The matching subdivision is located by applying a recovered motion vector to a co-located subdivision in the reference picture. The recovered motion vector is obtained using a reference motion vector from an available neighboring subdivision of the corrupted subdivision.