Image Segmentation Using Superpixel Seed Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image segmentation methods require cumbersome user inputs, such as scribbles, to separate foreground and background in digital images, which can be inefficient and prone to noise and visual ambiguities.

Innovation Solution

An image segmentation method that uses a seed generation engine to automatically generate foreground and background seeds by filtering and ranking superpixels based on geodesic distances and appearance metrics, reducing the need for extensive user input and improving segmentation accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If user-specified scribbles are used for foreground and background segmentation, then segmentation can be performed, but the process becomes cumbersome and requires extensive user input

Engineering Contradiction:
Improvesegmentation accuracyVSAvoiduser input complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs preliminary segmentation of the image into superpixels before user interaction. This pre-processing step organizes the image into meaningful regions, reducing the complexity of subsequent user input requirements while maintaining segmentation accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Superpixels serve as an intermediary between raw pixels and final segmentation regions. By introducing this intermediate representation, the system reduces the complexity of user input while preserving segmentation precision, as users interact with superpixel regions rather than individual pixels.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If traditional scribble-based methods are used, then segmentation is achieved, but the method is prone to noise and visual ambiguities

Engineering Contradiction:
Improvesegmentation robustnessVSAvoidnoise and visual ambiguities
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The image is segmented into superpixels, which are grouped regions of pixels with similar characteristics. This segmentation approach reduces noise and visual ambiguities by aggregating pixel-level variations into coherent regions, thereby improving segmentation reliability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system changes the parameter of representation from individual pixels to superpixels, and further to ranked superpixel sets. This parameter transformation filters out noise and ambiguities inherent in pixel-level data while preserving meaningful segmentation information.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If extensive user input is required for segmentation, then precise control is achieved, but productivity and efficiency are reduced

Engineering Contradiction:
Improvesegmentation controlVSAvoidsegmentation speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

Instead of requiring complete user specification of all foreground and background regions, the system uses partial user input (selective superpixel marking) combined with automatic algorithms to complete the segmentation. This reduces the amount of user action required while maintaining precision through the ranking and selection mechanism.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2911113B1Method, apparatus and computer program product for image segmentation
Publication Date: 2016.10.19 NOKIA TECHNOLOGIES OY
  • EP2911113B1 patent drawingFigure 1
  • EP2911113B1 patent drawingFigure 2
  • EP2911113B1 patent drawingFigure 3

AI summary

In an example embodiment, a method, apparatus and computer program product are provided. The method includes facilitating receipt of a plurality of superpixels of an input image and an object selection input for selecting part of a foreground object in the input image. The method includes determining first set of superpixels as first set of foreground seeds and second set of superpixels as first set of background seeds. The method includes filtering the first set of background seeds to generate second set of background seeds based on geodesic distances. The method includes adding superpixels of first set of unclassified superpixels to the second set of background seeds to generate third set of background seeds, and adding superpixels of second set of unclassified superpixels to the third set of background seeds to generate fourth set of background seeds based on applying a classifier on the second set of unclassified superpixels.