Image Sequence Encoding Using Pixel Trajectories for Better Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression methods do not effectively exploit the correlation between consecutive images, leading to suboptimal compression rates and storage efficiency, especially in applications like medical imaging where high volumes of data need to be stored and transmitted efficiently.
Innovation Solution
A method that encodes digitized images by predicting pixel values using trajectories of similar pixels from the same or adjacent images, with the prediction error encoded in a lossless or lossy manner, allowing for efficient compression and decoding without requiring additional side information for prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If individual images are compressed independently using standards like JPEG-LS or JPEG-2000, then lossless compression is achieved, but the compression rate is suboptimal because correlation between consecutive images is not exploited
Solution Approach 1:
The patent merges the compression of multiple consecutive images into a unified process, exploiting temporal correlations between images. Instead of compressing each image independently, the method combines information from multiple images to improve compression efficiency while maintaining lossless reconstruction.
Solution Approach 2:
The patent transitions from two-dimensional spatial compression (within a single image) to three-dimensional space-time compression by incorporating the temporal dimension. This allows exploitation of correlations across time (between consecutive images) in addition to spatial correlations, achieving better compression rates.
2Productivity
If video encoding standards like H.264/AVC are used with movement compensation, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent extracts and exploits the temporal correlation component from image sequences without implementing the full complexity of motion compensation algorithms. By focusing specifically on the correlation aspect rather than complete motion estimation and compensation, the method achieves good compression efficiency with reduced complexity.
3Reliability
If larger volumes of image data are stored and transmitted, then complete image sequences are preserved, but storage space and transmission bandwidth requirements increase
Solution Approach 1:
The patent merges information from multiple images by exploiting their correlations, allowing representation of complete image sequences with fewer bits. This combining approach maintains data completeness for reconstruction while reducing the total quantity of stored and transmitted data.
Data Source
AI summary
Pixels with associated pixel values form a sequence of digitized images that are encoded by predicting pixels of the images and encoding the prediction errors. At least a part of the pixels are encoded by determining trajectories, each running through a pixel to be encoded and through other previously encoded pixels from the image of the pixel being encoded and/or from one or more images that are temporally proximate to that image. For each of the trajectories determined, a valuation parameter is in inverse proportion to the deviations in the uncoded associated pixel values of the other pixels along a trajectory. A predicted pixel value of the pixel to be encoded is determined based on the other pixels of the trajectory with the highest valuation.