Neural Network Object Segmentation Without Depth Sensors

Solution Overview

Problem

Existing image analysis technologies require excessive calculations for depth information-based object segmentation, making them inefficient for extracting objects from images using only color information.

Innovation Solution

A method involving an image model, such as a neural network with nonlinear activation functions, is used to classify pixels based on attribute thresholds, generating a mask image and foreground image to segment objects from input images without relying on depth information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If depth information-based object segmentation technology is used, then object segmentation accuracy is improved, but device complexity and computational load increase due to requiring additional depth sensing modules and excessive calculations

Engineering Contradiction:

Improveobject segmentation accuracyVSAvoiddevice complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and removes the depth information processing component from the object segmentation system. Instead of using depth maps and additional depth sensors, the invention uses only color image information processed through a neural network model, thereby eliminating the complexity associated with depth sensing modules and depth information processing while maintaining segmentation functionality

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces the mechanical/optical depth sensing system with a computational approach using neural networks. Instead of physically measuring depth with additional sensors, the system uses image processing algorithms that analyze color information to achieve segmentation, substituting physical measurement mechanisms with computational models

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If depth information-based object segmentation technology is used, then object segmentation accuracy is improved, but processing time and computational resources increase due to excessive calculations for processing depth information

Engineering Contradiction:

Improveobject segmentation accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and removes the depth information processing component from the object segmentation system. Instead of using depth maps and additional depth sensors, the invention uses only color image information processed through a neural network model, thereby eliminating the complexity associated with depth sensing modules and depth information processing while maintaining segmentation functionality

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the input parameters from depth maps and color images to only color images. By transforming the segmentation approach to rely solely on color information processed through neural networks with activation functions, the system reduces computational complexity and processing time while maintaining segmentation accuracy

Inventive Principle:

Principle #35Parameter changes

3Device complexity

If color information-based object segmentation is used, then device complexity is reduced, but object segmentation accuracy deteriorates compared to depth information-based methods

Engineering Contradiction:

Improvedevice complexityVSAvoidobject segmentation accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent replaces the mechanical/optical depth sensing system with a computational approach using neural networks. Instead of physically measuring depth with additional sensors, the system uses image processing algorithms that analyze color information to achieve segmentation, substituting physical measurement mechanisms with computational models

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the input parameters from depth maps and color images to only color images. By transforming the segmentation approach to rely solely on color information processed through neural networks with activation functions, the system reduces computational complexity and processing time while maintaining segmentation accuracy

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10540768B2Apparatus and method to segment object from image

Publication Date: 2020.01.21 SAMSUNG ELECTRONICS CO LTD

AI summary

A method of segmenting an object from an image includes receiving an input image including an object; generating an output image corresponding to the object from the input image using an image model; and extracting an object image from the output image.