Unlock instant, AI-driven research and patent intelligence for your innovation.

Image Segmentation Method Combining Edge Detection and Watershed Algorithm

A watershed algorithm and image segmentation technology, applied in the field of fuzzy recognition, can solve the problems of inaccurate edges, unclosed edges, over-segmentation, and inconsistencies between edges and target boundaries, etc., and achieve fast results

Active Publication Date: 2019-04-26
WENZHOU UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, superpixels can only be used as a transition layer between the target and the pixel, and there are problems such as over-segmentation and inconsistency between the edge and the target boundary.
Without the support of the intermediate semantic layer, low-level pixels and high-level applications cannot be connected, and image segmentation has become an important reason hindering the further development of machine vision technology
[0003] In the long-term image segmentation research, it is found that the classic method Canny operator in image edge extraction has been able to obtain good results, but the Canny operator cannot find the correct gradient direction at the corner of the edge, and non-maximum suppression will occur. The phenomenon of losing the correct edge, so it is impossible to get a closed edge; in addition, the edge segmented by the Canny operator also has the defect of inaccuracy
Another classic algorithm in image segmentation, the watershed algorithm, can better extract object contours and accurately obtain object boundaries, but it is easily affected by noise, resulting in over-segmented images.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Image Segmentation Method Combining Edge Detection and Watershed Algorithm
  • Image Segmentation Method Combining Edge Detection and Watershed Algorithm
  • Image Segmentation Method Combining Edge Detection and Watershed Algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0015] Embodiment 1, the specific implementation steps of using templates of the same size are as follows: the first step, image edge extraction: use the Canny algorithm to extract the edge in the image; the second step, edge connection: for each Canny edge, use the same size template After expansion, the edges whose gap is less than 2 times the expansion diameter will be connected; the third step is to segment the image: use the area surrounded by Canny edges as the seed area of ​​the watershed algorithm, and then use the watershed algorithm to grow in the expanded area. The edges generated by the watershed algorithm segment the image into meaningful regions. For two regions with bottlenecks to connect, even if the two regions are split into two regions in the second step, they will still be connected into one region in the third step.

Embodiment 2

[0016] Embodiment 2, the specific implementation steps of using templates of different sizes are as follows: Step 1, use the contour detector to extract edges: Use the Canny algorithm to extract the edges in the image, and find all the endpoints and corners after single pixelation. The endpoint is a point with only one direction edge pixel in the 8-neighborhood, and the corner point is the point with three or more direction edge pixels in the 8-neighborhood. The corner point divides the edge into edge line segments (as shown in Figure 1, the corner represented by X points, triangles represent endpoints). The second step is to fill the gap: For each endpoint in the graph, fill the gap. For each endpoint of a Canny edge, find the nearest endpoint (let the distance to the nearest endpoint be Dpp). In order to prevent the nearest end point from being too far away, and there are close edges that can be connected, as in the case of end point 2 in Figure 2-a, then the edge line segm...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses an image segmentation method combining an edge detection algorithm with a watershed algorithm. The method comprises the following steps: step 1, extracting image edges: extracting the edges by adopting the edge detection algorithm; step 2, connecting the edges: performing expansion connection on the image edges extracted in the step 1 by adopting templates same in size, or connecting end points of the image edges extracted in the step 1, and then performing expansion by using templates different in size; and step 3, segmenting an image: performing growth by using the watershed algorithm in a growth region by taking an expansion region obtained in the step 2 as the growth region and other regions as seed regions of the watershed algorithm, confirming or cutting off the connection, and segmenting the image into regions. The method has the beneficial effects that the edge detection algorithm is utilized, so that a closed target profile is obtained and the position is more accurate than a Canny edge; and no problems of over-segmentation and the like of the watershed algorithm exist, the speed is high, and originally connected regions are not forcedly segmented.

Description

technical field [0001] The invention relates to an image segmentation method in the field of fuzzy recognition, in particular to an image segmentation method combining edge detection and watershed algorithm. Background technique [0002] Image segmentation represents an image as a collection of physically meaningful connected regions, that is, according to the prior knowledge of the target and background, the target and background in the image are marked and positioned, and then the target is separated from the background or other pseudo separated from the target. Image segmentation plays a connecting role in image understanding applications such as target recognition, object tracking, and behavior analysis, which greatly reduces the amount of data to be processed in subsequent image analysis and recognition processes, while retaining relevant image structure features Information. Image segmentation has been highly valued by people since the 1970s, and thousands of segment...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T7/13G06T7/11
Inventor 罗胜
Owner WENZHOU UNIV