Neural Network Image Extraction Using Multi-Stage Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for object extraction in images using convolutional neural networks (CNNs) are inefficient in terms of speed and accuracy.
Innovation Solution
An image processing apparatus and method that involves reducing an image, extracting a target structure, and inputting the reduced and corresponding images into a machine-learning neural network model to enhance the extraction of the target structure, utilizing a multi-channel input layer and convolutional processing layers for high-speed and accurate object extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CNN-based object extraction methods are used, then the extraction process can be performed, but the extraction speed is slow and accuracy is insufficient
Solution Approach 1:
The patent divides the object extraction process into multiple processing stages: initial extraction from the entire image, followed by focused extraction on specific regions of interest. This segmentation allows the system to first identify potential target areas quickly, then apply more intensive processing only where needed, thereby improving both speed and accuracy simultaneously
Solution Approach 2:
The patent introduces a spatial dimension by creating a region of interest map that highlights specific areas requiring detailed extraction. This additional dimensional layer guides the neural network to focus computational resources on critical regions rather than uniformly processing the entire image, resolving the contradiction between comprehensive accuracy and processing speed
2Measurement precision
If the neural network processes the entire image at full resolution, then extraction accuracy may be maintained, but processing time increases significantly
Solution Approach 1:
The patent extracts and isolates regions of interest from the full image, creating a focused subset of data for detailed processing. By taking out only the relevant portions that contain potential target objects, the system maintains extraction accuracy while dramatically reducing the computational burden and processing time required
Solution Approach 2:
The patent applies partial processing to the entire image by identifying and concentrating computational efforts on specific regions rather than uniformly processing all pixels. This selective approach ensures sufficient accuracy for target detection while avoiding the excessive processing time that would result from analyzing the complete high-resolution image in detail
Data Source
AI summary
A processor is configured to: reduce a target image to derive a reduced image; extract a region of a target structure from the reduced image to derive a reduced structure image including the region of the target structure; extract a corresponding image corresponding to the reduced structure image from the target image; and input the corresponding image and the reduced structure image into an extraction model constructed by machine-learning a neural network to extract a region of the target structure included in the corresponding image from the extraction model.