Intra-Divided Picture Encoding for Reduced Computational Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional intra-frame predictive coding methods in H.264 face challenges in reducing decoding computational complexity while maintaining coding efficiency, especially for images with non-repeating patterns, as they rely on deblocking filters and motion vector searches, which increase processing complexity and do not effectively reduce prediction errors.
Innovation Solution
The method involves dividing input pictures into sub-blocks, performing intra-frame encoding on some sub-blocks and inter-frame predictive encoding using encoded sub-blocks as reference pictures, with adaptive selection based on distance and quantization precision, allowing for flexible reference picture selection and combination of predicted pictures to enhance prediction efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deblocking filter is applied to entire frame to suppress block distortion, then prediction efficiency is improved, but computational complexity increases
Solution Approach 1:
The frame is divided into multiple blocks, and only specific blocks (e.g., blocks with large prediction errors or at block boundaries) are processed by the deblocking filter, rather than applying the filter to the entire frame. This selective processing reduces computational complexity while maintaining prediction efficiency where it is most needed.
Solution Approach 2:
The deblocking filter is applied with different strengths or only in specific local regions where block distortion is most prominent, such as at block boundaries or in areas with high prediction error. This localized application improves prediction efficiency in critical areas without unnecessarily increasing computational complexity across the entire frame.
2Measurement precision
If motion vector search is performed to find reference block, then prediction accuracy is improved, but decoding computational complexity increases
Solution Approach 1:
Motion vectors or reference block positions are determined during the encoding phase and stored in the bitstream. During decoding, the decoder directly uses these pre-determined motion vectors without performing exhaustive motion vector searches, thereby maintaining prediction accuracy while significantly reducing decoding computational complexity.
Solution Approach 2:
Instead of performing motion vector searches during decoding, the decoder copies reference block information from previously decoded blocks or from encoder-side calculations that are transmitted in the bitstream. This copying approach maintains prediction accuracy while avoiding the computational burden of repeated searches.
3Measurement precision
If block-matching based intra-frame prediction is used to reduce prediction error, then coding efficiency is improved, but it is ineffective for pictures without repeating patterns and computational complexity remains large
Solution Approach 1:
The prediction method is made adaptive by dynamically selecting between block-matching based prediction and other prediction methods (such as gradient-based or boundary-based prediction) depending on the characteristics of the current block. This dynamic selection allows the system to achieve good prediction error reduction for both repeating and non-repeating patterns while avoiding the computational complexity of block-matching when it is not applicable.
Solution Approach 2:
The prediction approach is changed based on block characteristics such as variance, gradient, or pattern detection. When repeating patterns are detected, block-matching is applied; when non-repeating patterns are detected, alternative prediction methods with lower computational complexity are used. This parameter-based adaptation maintains prediction accuracy across different image content types while reducing overall computational complexity.
Data Source
Figure 1
Figure 2A~2B
Figure 2C~2D
AI summary
The encoding computational complexity and the decoding computational complexity are reduced while suppressing deterioration of coding efficiency in comparison with conventional intra divided-picture coding. A divided picture generation unit divides an input encoding target picture into blocks having the same size, and collects pixels having the same relative position in the blocks to generate divided pictures having the same size. An intra divided-picture encoding unit performs intra divided-picture encoding on some divided pictures. A reference picture selection unit selects, as a reference picture, an encoded divided picture in which a pixel position on an original picture is close in order to encode the other encoding target divided pictures. An inter divided-picture predictive encoding unit generates a predicted picture using the selected reference picture and encodes an encoding target divided picture through inter divided-picture prediction in units of divided pictures.