Segmented Image Compression With Class-Specific ROI Encoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image and video compression methods fail to optimize encoding for specific regions of interest in smart surveillance and machine vision tasks, leading to suboptimal compression efficiency and performance in tasks like object classification and recognition.

Innovation Solution

An image data compression method that segments images into regions, applies class-specific learning-based encoders, and multiplexes their outputs, optimizing each encoder for the characteristics of the identified regions, including agnostic and semantic content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If traditional compression algorithms (JPEG, JPEG2000) are used, then compression speed is acceptable, but compression ratio is low and block effect is obvious

Engineering Contradiction:
Improvecompression ratioVSAvoidalgorithm complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The image is divided into multiple blocks, and each block is further segmented into multiple regions based on gradient direction. This multi-level segmentation allows the algorithm to process different regions with appropriate compression strategies, achieving high compression ratios while maintaining image quality by preserving important edge information.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different compression strategies are applied to different regions within the image blocks. Regions with important edge information (identified by gradient analysis) are preserved with higher quality, while smooth regions are compressed more aggressively. This local quality approach ensures high compression ratios without obvious block effects.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If image quality is prioritized over compression speed, then image quality is good, but compression speed is slow

Engineering Contradiction:
Improveimage qualityVSAvoidcompression speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The algorithm performs preliminary gradient calculation and region classification before compression. By pre-identifying edge regions and smooth regions, the algorithm can quickly apply appropriate compression strategies during the main compression phase, maintaining high image quality while improving compression speed through efficient preprocessing.

Inventive Principle:
Principle #10Preliminary action

3Loss of information

If compression ratio is increased, then less data is transmitted, but block effect becomes more obvious

Engineering Contradiction:
Improvedata transmission volumeVSAvoidblock effect
Core Design Contradiction:
Loss of informationVSObject-affected harmful factors

Solution Approach 1:

The algorithm transforms image data from spatial domain to frequency domain using DCT, and applies quantization with different parameters for different regions. By changing the representation parameters and applying region-specific quantization matrices, the algorithm achieves high compression ratios while minimizing visible block effects through perceptual optimization.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4511816B1Image data compression method and device using segmentation and classification
Publication Date: 2026.04.29 INST DE TELECOMUNICACOES
  • EP4511816B1 patent drawingFigure 1
  • EP4511816B1 patent drawingFigure 2A~2C
  • EP4511816B1 patent drawingFigure 3

AI summary

Device and method for image data compression using segmentation and classification, the method comprising the steps of: identifying regions in a received image comprised of image pixels; segmenting the image pixels into segmented regions, each segmented region corresponding to an identified region, and into an image background comprised of image pixels, if existing, not belonging to any of the identified regions; determining a class for each segmented image region from a plurality of predetermined image classification classes; applying an image learning-based encoder to each segmented image region, according to the determined class of each segmented image region, wherein a specific image learning- based encoder has been preselected for each of the image classification classes from a library of image learning-based encoders; outputting the encoded segmented image regions.