Image Data Stream Coding With Constant-Value Block Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression techniques, such as HEVC, are not optimal for content with strong discontinuities like screenshots or synthesis images, as they struggle to accurately predict pixel values across transitions.
Innovation Solution
A method that determines a group of constant pixel values within a block and uses these values to predict pixels in transition areas, reducing the need for additional information transmission by selecting a prediction mode based on the distance between the original pixel value and the predicted value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional prediction techniques (intra prediction using neighbouring pixels) are used, then natural images can be compressed effectively, but images with strong discontinuities (screenshots, synthesis images) suffer from high prediction error
Solution Approach 1:
The prediction mode is dynamically selected based on the characteristics of the current block. The encoder determines whether to use conventional intra prediction or the improved method by evaluating the block's properties, allowing the system to adapt between different prediction strategies depending on the image content type.
Solution Approach 2:
The invention changes the prediction parameter by introducing a threshold value for the absolute difference between pixel values. When this threshold is exceeded, the prediction method switches from using neighbouring pixel values to using constant values from a predefined list, thereby adapting to discontinuous content.
2Loss of energy
If prediction is performed using neighbouring pixels, then smooth transitions are captured well, but strong discontinuities (sharp transitions between distinct levels) generate high residual energy
Solution Approach 1:
Instead of computing prediction values from neighbouring pixels, the invention copies predefined constant values from a list into the prediction block. These constant values represent the distinct levels present in the image and are directly copied to replace pixel values in transition areas, eliminating the need for complex prediction calculations.
Solution Approach 2:
The invention segments the prediction process into two distinct approaches: conventional intra prediction for smooth regions and constant value prediction for discontinuous regions. The segmentation is achieved by dividing the image into blocks and applying different prediction methods to different blocks based on their characteristics.
3Productivity
If a sophisticated prediction method is used to handle discontinuities, then compression performance improves, but the complexity of the coding system increases
Solution Approach 1:
The constant values used for prediction are prepared in advance and stored in a predefined list during the encoding process. This preliminary action eliminates the need for complex real-time calculations when handling discontinuities, as the encoder only needs to select from pre-computed constant values rather than performing sophisticated prediction algorithms during encoding.
Solution Approach 2:
The prediction system serves itself by automatically determining which prediction method to use based on simple threshold evaluation. The encoder assesses the block characteristics and self-selects the appropriate prediction strategy without requiring external intervention or complex control logic, thereby maintaining simplicity while improving performance.
Data Source
AI summary
A method and device for coding and decoding a data stream representing an image split into blocks. For a current block, a group of pixel values in the block is determined from previously decoded pixels, and for each pixel, a quantized prediction residue, in the spatial domain, is decoded. A prediction value for the pixel is determined according to a first prediction mode by predicting the pixel from at least one other previously decoded pixel of the current block. Information is decoded from the stream indicating whether the pixel is predicted according to a second mode using a prediction resulting from the group of pixel values in the block. When the pixel is predicted according the second mode, the prediction value for the pixel is replaced with a selected value of the group. The pixel is reconstructed using the prediction value associated with the pixel and the de-quantized prediction residue.