Iterative Foreground Detection in Image Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for detecting salient foreground objects in images struggle with multiple objects in a scene, particularly failing to accurately detect and segment multiple disconnected objects.

Innovation Solution

An iterative approach is employed, incorporating a modified background prior and high-dimensional embedding to enhance foreground segmentation, using the Image Region Adjacency Graph and eigenvectors to compute saliency scores, along with a Silhouette score and adaptive thresholding to determine the ideal number of eigenvectors and iterations for accurate multi-object segmentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a single Fiedler vector is used for salient foreground detection, then the computation is efficient and simple, but the method can only detect a single salient object and fails to handle multiple disconnected objects

Engineering Contradiction:
Improvecomputation efficiencyVSAvoidability to detect multiple objects
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent applies segmentation by dividing the detection process into multiple iterations, where each iteration segments out one salient object from the image. The Image RAG is repeatedly segmented by computing Fiedler vectors, with each iteration isolating one foreground object, thereby enabling detection of multiple disconnected objects while maintaining computational efficiency through iterative simplification of the graph structure

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from a single-dimension approach (one Fiedler vector) to a multi-dimensional approach by computing multiple Fiedler vectors across iterative dimensions. Each iteration operates in a reduced-dimensional space after removing previously detected objects, allowing the method to capture multiple salient objects that exist in different dimensional spaces of the Image RAG

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Productivity

If the background prior is simplified, then the computation is faster, but the foreground segmentation accuracy decreases

Engineering Contradiction:
Improvecomputation speedVSAvoidforeground segmentation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing the background prior information from the Image RAG before the iterative segmentation process begins. The background node and its associated prior information are prepared in advance, allowing each iteration to quickly reference this pre-established background model without re-computing it, thus maintaining both speed and accuracy

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3559906B1An iterative method for salient foreground detection and multi-object segmentation
Publication Date: 2024.02.21 KODAK ALARIS INC
  • EP3559906B1 patent drawingFigure 1
  • EP3559906B1 patent drawingFigure 2
  • EP3559906B1 patent drawingFigure 3

AI summary

A system and method that performs iterative foreground detection and multi-object segmentation in an image is disclosed herein. A new background prior is introduced to improve the foreground segmentation results. Three complimentary methods detect and segment foregrounds containing multiple objects. The first method performs an iterative segmentation of the image to pull out the salient objects in the image. In a second method, a higher dimensional embedding of the image graph is used to estimate the saliency score and extract multiple salient objects. A third method uses a metric to automatically pick the number of eigenvectors to consider in an alternative method to iteratively compute the image saliency map. Experimental results show that these methods succeed in accurately extracting multiple foreground objects from an image.