Research On PCB CT Image Segmentation Based On Level Set Method

It is a new technique to non-destructively test(NDT)the printed circuit board(PCB)through computed tomography(CT),which has some advantages such as non-destructive,efficient and high-definition.Image segmentation is a key step for NDT because it will have great influence on automatically finding the position of faults.Therefore,it is significant to study efficient method to segment CT images of PCB.Recently,image segmentation based on level set has been widely studied.This method represents the curve through level set function in higher dimension space,which makes it to adapt the change of topology and to extend three-dimension easily.However,the CT images of PCB have some drawbacks for image segmentation based on level set like intensity inhomogeneity and large scale.Aiming at these problems,some improved methods have been proposed in this work as follows:1.This study proposes an improved level set method for image segmentation based on local variance and improved intensity inhomogeneity model to solve the problem of intensity inhomogeneity in PCB images.The modified model considers the difference between an original image and an estimated image without bias field,which can obtain better segmentation than traditional one.Apart from using this difference,Gaussian distribution with means and variance is utilized as the local intensity descriptor to map the original image into another domain so the object and the background can be better separated in the transformed domain.Then,an improved level set energy function that combines the image term,local variance,and the above difference is defined.The minimization of the function can be processed by level set evolution.The proposed method is compared with existing methods,and experiments on both synthetic and real images demonstrate that our method has superior performance.2.Aiming at the seriously low contrast of intensity in PCB images caused by artifacts,a novel level set method based on shape prior is proposed.Firstly,PCB CT image is filtered by different orientation Gabor filters according to the obvious orientation feature of lines in PCB.The shape prior can be obtained after localization method.Then,the energy function consists of Chan-Vese item,localized energy item and shape prior item which is represented by probabilistic definition of shape.Finally,the segmentation result can be obtained by minimizing the energy function.The experiments of PCB CT images with low contrast of intensity demonstrate the performance of our model.The efficiency of our model has better performance as well.3.Level set method for image segmentation has poor performance on efficiency because of the large scale of PCB images.In this work,we propose to use continuous max-flow algorithm to optimize a locally improved Chan-Vese model for image segmentation in the presence of intensity inhomogeneity.After analyzing the problem of continuous max-flow,we discuss the similarity between this problem and the energy function of local Chan-Vese model.Then,we convert this energy function to the frame of Graph Cut whose energy function can be efficiently minimized by max-flow algorithm.As a result,the process of optimization of local Chan-Vese model can be accelerated by using max-flow algorithm.The experiments demonstrate that the proposed method can achieve satisfactory segmentation for images with intensity inhomogeneity as well as very high efficiency.For the large scale PCB images,this method also has high efficiency and it provides a possible method for segmentation of PCB images.