基于双重Sigmoid变换的PCB图像增强算法

PCB Image Enhancement Algorithm Based on Double Sigmoid Transform

在基于边缘的PCB缺陷检测中提出了一种新的图像变换方法——双重Sigmoid变换,并导出了Sigmoid变换算子。试验表明,采用该变换对PCB图像进行增强能有效地锐化边缘,对边缘附近的噪声滤除能力强,有利于精确定位边缘,对原图像的清晰度及均匀性要求不高。

In the Printed Circuit Board （PCB） defect detection based on Automatic Optic Inspection （AOI）, the actually collected PCB images are often blurred over the edge, especially in the area of dense vertical lines. Therefore, to eliminate noise and enhance the image to highlight the edge, which makes the alignment between PCB sample and the detected PCB easer and more precise and avoid missing and false detection, are needed. The main target for enhancing PCB images is to sharp the edges, at the same time to effectively suppress noise, especially the noise near the edges. From the perspective of the edge sharpening, the differential operation should be used. However, for filtering out the noise, integral operation should be applied to. Based on the above features, a new image transformation i.e. double Sigmoid transformation, which imposes a double transformation on the original image and at the same time uses the results with the transformation to enhance the original image, is presented. At first, a definition of Sigmoid transformation is given, and then double Sigmoid operators are deduced. In order to simplify the computing, only a set of special feature value is taken to calculate the sigmoid operator, which makes the computing speed faster. Experiments show that the Sigmoid algorithm could effectively sharpen PCB image edges, at same time, filter out noise in the image. Even if the original PCB image is not distinct enough and uniform brightness, the algorithm is still able to precisely locate circuit edge, which could make the edge extracting and edge recognition easier.