Image Block Fingerprinting for Accurate ROI Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image segmentation techniques are tedious and prone to user error, requiring manual input of boundary lines and areas, which affects accuracy and efficiency.
Innovation Solution
An image processing system that segments images by dividing them into blocks, generating a compressed representation or 'fingerprint' using a perceptual hash algorithm, grouping blocks by similarity, and extracting regions of interest based on these categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models are used for image segmentation, then segmentation accuracy is improved, but computational resources and training time are increased
Solution Approach 1:
The patent divides the image segmentation task into multiple stages: a coarse segmentation stage using a lightweight model and a refinement stage using a deeper model only on relevant regions. This segmentation of the processing task reduces overall computational resources while maintaining high accuracy.
Solution Approach 2:
The patent applies partial action by using a deeper segmentation model only on regions identified as relevant by the coarse model, rather than applying the deep model to the entire image. This reduces computational expenditure while maintaining high accuracy where needed.
2Adaptability or versatility
If existing segmentation models are used, then segmentation capability is provided, but they fail to segment objects with similar appearances and cannot handle 3D objects effectively
Solution Approach 1:
The patent introduces dynamic 3D bounding box representations that can adapt to different object orientations and configurations. The system dynamically adjusts the 3D box parameters based on the input image, enabling reliable segmentation of objects with similar appearances and complex 3D structures.
Solution Approach 2:
The patent transitions from 2D image segmentation to 3D object representation by introducing depth information and volumetric bounding boxes. This dimensional extension enables the system to handle 3D objects and distinguish between objects with similar 2D appearances but different 3D structures.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method for image segmentation. The system may include at least one processor that is directed to obtain an image (502); divide the image into a plurality of image blocks (504); determine, for each of the plurality of image blocks, a compressed representation of one or more features of the image block (506); group the plurality of image blocks into at least two different categories based on the plurality of compressed representations (508); and extract a region of interest from the image based on at least one of the at least two different categories of the plurality of image blocks (510).