Semi-automatic Image Labeling with Iterative Grabcut and Edge Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing grabcut image cutting algorithm faces challenges in achieving high precision when segmenting foreground images with complex contour shapes, leading to inclusion of background pixels and reduced accuracy.
Innovation Solution
A semi-automatic image data labeling method is introduced, which executes the grabcut algorithm multiple times to acquire the outline coordinates of a foreground image, and employs an edge tracking algorithm to update local coordinate sets, thereby improving labeling data precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the grabcut algorithm is used with a single frame drawn on the foreground image, then the operation is simple and fast, but the frame range becomes very large for images with special contour shapes, resulting in many background pixels and decreased segmentation precision
Solution Approach 1:
The patent divides the segmentation process into multiple iterative stages. Instead of drawing one large frame, the system performs multiple grabcut algorithms with progressively refined frames, segmenting the foreground more precisely in each iteration until the desired precision is achieved.
Solution Approach 2:
The patent introduces dynamic adjustment of the frame range through iterative optimization. The frame is dynamically refined in each iteration based on the previous segmentation result, allowing the system to adapt to complex contour shapes while maintaining operational simplicity through automation.
2Measurement precision
If multiple frames are drawn to improve segmentation precision for complex shapes, then the segmentation accuracy improves, but the operation complexity and time consumption increase
Solution Approach 1:
The patent implements automatic frame optimization where the system self-adjusts the frame range in each iteration based on the previous segmentation result. The automatic frame optimization module refines the frame without requiring manual redrawing, allowing multiple iterations while keeping operation complexity low.
Solution Approach 2:
The patent uses feedback from previous segmentation results to guide subsequent iterations. The system analyzes the segmentation outcome and automatically adjusts the frame for the next iteration, creating a closed-loop process that improves precision while minimizing manual intervention.
3Measurement precision
If multiple iterations of the grabcut algorithm are performed, then the labeling data precision improves, but the time consumption increases
Solution Approach 1:
The patent applies partial action by performing the full iterative process only when necessary. The system can stop after one iteration if the initial segmentation is sufficient, or continue with additional iterations only when complex shapes require higher precision, thus balancing time consumption with precision requirements.
Solution Approach 2:
The patent performs preliminary assessment of the image characteristics to determine the appropriate number of iterations needed. By analyzing the complexity of the foreground shape and background patterns before full segmentation, the system can pre-determine the optimal iteration count, avoiding unnecessary time consumption while ensuring sufficient precision.
Data Source
AI summary
Disclosed are a semi-automatic image data labeling method, an electronic apparatus and a non-transitory computer-readable storage medium. The semi-automatic image data labeling method may include: displaying a to-be-labeled image, the to-be-labeled image comprising a selected area and an unselected area; acquiring a coordinate point of the unselected area and a first range value; executing a grabcut algorithm based on the coordinate point of the unselected area and the first range value acquired, and obtaining a binarized image divided by the grabcut algorithm; executing an edge tracking algorithm on the binarized image to acquire current edge coordinates; updating a local coordinate set based on the current edge coordinates acquired; updating the selected area of the to-be-labeled image based on the local coordinate set acquired.