Adaptive Video Transform and Prediction Block Sizing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video signal processing methods fail to efficiently eliminate spatial redundancy and achieve optimal compression ratios, particularly in intra-picture prediction and block unit encoding/decoding, due to limitations in transform sizes and prediction block sizes within macroblocks.
Innovation Solution
The method involves performing DC or Hadamard transforms on macroblocks larger than 16x16, using flag information to indicate transform sizes, and varying prediction block sizes within a macroblock for intra-picture prediction, allowing for adaptive processing and efficient data compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If DC transform or Hadamard transform is performed on macroblocks larger than 16x16 with various transform sizes, then spatial redundancy is eliminated and compression efficiency is improved, but device complexity and processing overhead increase
Solution Approach 1:
The macroblock is divided into multiple blocks of different sizes (e.g., 16x16, 16x8, 8x16, 8x8, 4x4) to which different transform sizes are applied. This segmentation allows the patent to eliminate spatial redundancy more effectively by matching transform sizes to actual content characteristics in different regions, while the modular structure manages complexity through standardized processing units for each block size.
Solution Approach 2:
The patent dynamically selects transform sizes based on content characteristics and prediction block sizes. The transform size is not fixed but adapts to the local image content and prediction mode, allowing the system to optimize compression efficiency for each block while maintaining manageable complexity through conditional selection logic rather than exhaustive processing.
2Measurement precision
If various prediction block sizes are used within one macroblock for intra-picture prediction, then prediction accuracy is enhanced and residual data size is reduced, but processing complexity increases
Solution Approach 1:
Different prediction block sizes (e.g., 16x16, 16x8, 8x16, 8x8, 4x4) are applied to different regions within a macroblock based on local content characteristics. This allows each region to use the prediction block size most suitable for its specific features, enhancing overall prediction accuracy while maintaining complexity through localized rather than global processing.
Solution Approach 2:
The macroblock is segmented into multiple prediction blocks of varying sizes, each processed independently with appropriate prediction modes. This segmentation enables the system to capture local variations in the image content more accurately while managing complexity by processing each segment with standardized algorithms tailored to its size.
3Loss of information
If flag information indicating transform size is used, then data transfer size is reduced and encoding efficiency is improved, but decoding complexity increases
Solution Approach 1:
The patent extracts and transmits only the essential transform size information through compact flag fields in the bitstream, rather than transmitting complete transform parameters. This extraction approach significantly reduces the data transfer size while the decoder, equipped with knowledge of the standardized flag encoding schemes, can efficiently reconstruct the transform size information without excessive complexity.
Data Source
AI summary
A method of processing a video signal is disclosed. The present invention includes obtaining a DC (discrete cosine) transform coefficient for a current macroblock and partition information of a DC (direct current) component block from a bitstream, obtaining transform size information of each partition of the DC component block based on the partition information of the DC component block, performing an inverse DC transform or a Hadamard transform based on the transform size information of the partition of the DC component block, performing inverse quantization on a result value from the transform and an AC (alternating current) component, and reconstructing a residual block by performing an inverse DC transform on a result value from the inverse quantization, wherein the residual block includes a block indicating a pixel value difference between an original picture and a predicted picture.