Integer Image Compression Using Sub-0.5 Floating-Point Error Bounds

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The growing amount of data generated daily poses challenges in storage and transmission due to the gap between compute bandwidth and input/output access bandwidth, necessitating efficient data compression techniques to reduce storage, retrieval, and transmission bandwidths.

Innovation Solution

Implementing a method for compressing and decompressing integer values using a lossy floating-point compression scheme with an error bound of less than 0.5, allowing for high compression ratios while maintaining data fidelity by rounding floating-point pixel values to nearest integers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If lossless integer compression schemes are used, then data fidelity is maintained, but compression ratios are low

Engineering Contradiction:

Improvedata fidelityVSAvoidcompression ratio

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent changes the numerical parameter representation from integer to floating-point format, allowing the use of error bounds to control precision. By setting an error bound of less than 0.5, the scheme ensures that rounded values reconstruct the original integer losslessly while achieving higher compression ratios through more efficient encoding of floating-point values.

Inventive Principle:

Principle #35Parameter changes

2Quantity of substance

If data compression is increased to reduce storage and bandwidth, then storage costs and bandwidth requirements are reduced, but data reconstruction accuracy may be compromised

Engineering Contradiction:

Improvestorage bandwidthVSAvoiddata reconstruction accuracy

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent implements a feedback mechanism where the error bound parameter (less than 0.5) is used to control the compression process. This feedback ensures that the compression algorithm adjusts its precision to maintain lossless reconstruction, balancing compression efficiency with data accuracy requirements.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

By changing from integer to floating-point representation with controlled error bounds, the patent achieves higher compression ratios while maintaining the ability to perfectly reconstruct original integer values through rounding, thus resolving the trade-off between compression and accuracy.

Inventive Principle:

Principle #35Parameter changes

3Quantity of substance

If the gap between compute bandwidth and I/O access bandwidth is addressed through compression, then I/O bandwidth requirements are reduced, but compression and decompression processing time increases

Engineering Contradiction:

ImproveI/O bandwidthVSAvoidprocessing time

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The patent changes the data representation to floating-point with error bounds, which enables more compact encoding that reduces I/O bandwidth requirements. The simple rounding operation for decompression minimizes processing time, effectively addressing the bandwidth gap without excessive processing overhead.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11854235B1Lossless integer compression scheme

Publication Date: 2023.12.26 MARVELL ASIA PTE LTD

AI summary

Decompressing a compressed image to obtain a decompressed image includes receiving, in a compressed stream, compressed pixel values of the compressed image; decompressing, from the compressed stream, a first compressed pixel value of the compressed pixel values using a lossy floating-point decompression scheme to obtain a floating-point pixel value; rounding the floating-point pixel value to a nearest integer to obtain a pixel value of the decompressed image; and displaying or storing the decompressed image.