Lossless Sensor Data Compression Using Slices and Median Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression technologies often provide lossy compression, and there is a need for near real-time lossless compression techniques that are not asymmetric, allowing for faster encoding speeds.
Innovation Solution
Sensor data is divided into component planes and slices, with each sample encoded using a median-based prediction and context-driven model to compute and encode errors, utilizing symmetric encoding and decoding algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If lossless compression techniques are used, then compression quality is improved, but encoding speed deteriorates due to asymmetric encoding algorithms being multiple times slower than decoders
Solution Approach 1:
The patent segments the data processing into distinct planes (luma and chroma) and further divides each plane into multiple slices. This segmentation allows parallel processing of different slices during encoding, significantly improving encoding speed while maintaining lossless compression quality. Each slice can be processed independently by different processing units, eliminating the bottleneck of sequential processing in traditional asymmetric algorithms.
Solution Approach 2:
The patent inverts the traditional asymmetric approach by designing a symmetric compression algorithm where the encoder and decoder have comparable computational complexity. This is achieved by using simple prediction modes and context models that require similar processing power on both encoding and decoding sides, thereby improving encoding speed without sacrificing compression performance.
2Loss of information
If traditional asymmetric encoding algorithms are used, then lossless compression is achieved, but system resource consumption increases due to the complexity of the encoder
Solution Approach 1:
By dividing the data into multiple slices that can be processed independently, the patent reduces the computational burden on any single processing unit. Each processing unit handles only a portion of the data, reducing memory requirements and computational complexity while maintaining overall compression quality through coordinated processing of all slices.
Solution Approach 2:
The patent applies different context models to different regions (slices) of the data based on local characteristics. Each slice can use context models optimized for its specific content type, improving compression efficiency without requiring a single overly complex encoder that must handle all possible data patterns uniformly.
3Loss of information
If lossless compression is implemented, then data fidelity is improved, but processing time increases making real-time compression difficult
Solution Approach 1:
The patent divides the data stream into multiple parallelizable slices that can be processed simultaneously. This parallelization dramatically reduces processing time by utilizing multiple processing units working concurrently on different slices, enabling real-time lossless compression while maintaining complete data fidelity through coordinated reconstruction of all slices.
Solution Approach 2:
The patent performs preliminary organization of data into slices and prepares context models in advance before the actual compression processing. This preliminary structuring enables more efficient processing during the compression phase, reducing overall processing time while ensuring that all necessary information is available for lossless reconstruction.
Data Source
AI summary
Compression techniques are described. In an embodiment, a first plane of sensor data is accessed, the first plane of sensor data is divided into a plurality of slices, each sample is encoded in each slice from the plurality of slices, where encoding a sample include computing a median based prediction for the sample, computing an error for the sample comprising a difference between the sample and the computed median based prediction, determining a context for the sample, selecting a model for the sample by using the determined context, and encoding the computed error by using the selected model.