An image segmentation method combining global and local information level sets

Abstract

The invention discloses an image segmentation method which fuses global and local information level sets. The method provides a segmentation model, comprising a global item, a local item and a modified item. The global term is based on the classical CV model. The local term is based on the piecewise constant assumption. For each point in the region, the kernel function of the mean gray value of the point region and the gray value of other points in the neighborhood is used to define the local term. In the dynamic adjustment process, the model weakens the global term CV model based on the assumption that the gray level is uniform, but also weakens the weak ability of the local term to weak edge recognition. Modifications consist of smoothing constraints and penalties to avoid reinitialization and reduce computational time. The level set method can describe the evolution of curves implicitly on Cartesian mesh, avoid the problem that many curves can not be described by parametric description curves, and at the same time, it is easy to calculate the geometrical features of closed contours, such as curvature, unit normal vectors and so on.

Description

technical field [0001] The invention relates to the field of image segmentation research, in particular to an image segmentation method for integrating global and local information level sets. Background technique [0002] High-density flexible printed circuits are widely used in computers, medical care, transportation, and military industries. Due to the small wire width and spacing, high wiring density and precision, defects such as short circuits, open circuits, voids, protrusions, and depressions are prone to occur during the production process, which affects the performance of high-density flexible printed circuits. In the defect detection of high-density flexible printed circuits, the high-density flexible printed circuit images collected by industrial cameras are used as the application object. From image preprocessing to defect detection algorithms, it is necessary to provide more accurate segmented images to be more accurate. detection defect problem. From this we...

AI Technical Summary

Problems solved by technology

However, in the existing segmentation model based on global level set, it is assumed that the image to be segmented is a uniform gray image, while the high-density flexible printed circuit image is an uneven gray image, which makes the segmentation effect inaccurate; the segmentation model based on local level set In , the contour line is easy to fall into the local minimum, resulting in over-segmentation

Images