Composite Image Segmentation via Geometric Line Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for segmenting composite images into individual images are not robust for backgrounds with varied colors and textures, and require assumptions about the number and size of images, making them inefficient and user-frustrating.

Innovation Solution

A method that uses a data processor to locate line segments, define candidate rectangles, and select enclosing rectangles based on geometric constraints to segment composite images, regardless of background color or texture, and the number and size of individual images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If automatic photo segmentation algorithms assume white or black background and identify rectangles by content mismatch, then the segmentation can be performed automatically, but the algorithm is not robust for composite images having backgrounds with other colors and textures

Engineering Contradiction:

Improveautomatic segmentationVSAvoidrobustness to background variation

Core Design Contradiction:

Extent of automationVSReliability

Solution Approach 1:

The algorithm changes from assuming fixed background colors (white/black) to detecting line segments based on geometric properties and color consistency. It identifies rectangles by finding parallel and perpendicular line segments with consistent colors, rather than by content mismatch against expected background regions.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces the content-mismatch detection mechanism with a line-segment-based geometric detection mechanism. Instead of comparing pixel content against expected background, the system detects edges as line segments and uses their geometric relationships (parallelism, perpendicularity) to define rectangles.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Device complexity

If automatic photo segmentation algorithms make assumptions about the number and size of individual images, then the processing can be simplified, but the algorithm cannot handle composite images with varying numbers and sizes of images

Engineering Contradiction:

Improvealgorithm complexityVSAvoidadaptability to different image configurations

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The algorithm dynamically adapts to the actual number and arrangement of images in the composite image. Instead of assuming a fixed number of images, it detects all line segments and forms rectangles based on geometric constraints, allowing the output to naturally match the actual image configuration regardless of how many individual images are present.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent segments the composite image detection process into independent line segment detection, rectangle formation from line segments, and final image identification stages. This modular approach allows each stage to operate independently without assumptions about the final number or size of images.

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If users manually select rectangular areas for scanning, then precise control over scan regions is achieved, but users must perform multiple scan operations to scan all individual images

Engineering Contradiction:

Improvescan region precisionVSAvoidtime for multiple scan operations

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs self-service by automatically detecting and segmenting all individual images in the composite image without requiring user intervention. The algorithm identifies line segments, forms candidate rectangles, applies geometric constraints, and extracts individual images autonomously, eliminating the need for multiple manual scan operations.

Inventive Principle:

Principle #25Self-service

4Manufacturing precision

If image editing applications are used to manually select and save individual images, then precise segmentation control is achieved, but the process requires special software and is very time-consuming

Engineering Contradiction:

Improvesegmentation accuracyVSAvoidtime for manual selection and saving

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent replaces manual interactive selection with automated computational detection. The system uses line segment detection, geometric constraint satisfaction, and rectangle formation algorithms to automatically identify and extract individual images, eliminating the need for special image editing software and manual operations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS8467606B2Method for segmenting a composite image

Publication Date: 2013.06.18 KODAK ALARIS LLC

AI summary

A method for segmenting a composite image containing a plurality of individual images, comprising: locating line segments corresponding to edges in a composite image; defining a plurality of candidate rectangles by identifying sets of four line segments having a first pair of line segments that are parallel to and a second pair of line segments that are perpendicular to the first pair of line segments; identifying image regions that are more than a predefined isolation distance from any of the line segments; identifying a set of enclosing rectangles from among the candidate rectangles for each of the identified image regions; selecting a rectangle for each of the identified image regions from the corresponding set of enclosing rectangles by applying geometric constraints; forming a plurality of sub-images from the composite image by extracting the digital image data corresponding to the selected rectangles; and; and storing the sub-images in a processor-accessible memory.