A thresholding method for circuit pattern images containing low contrast signals

Abstract

This paper describes a method for thresholding gray-scale images to produce binary images in real time. It is specifically designed for thresholding images that contain both small, low-contrast defect patterns and relatively large, high-contrast circuit patterns without image deformation. This method analyzes peaks and troughs in the detected signal waveforms to threshold the low-contrast patterns ; it applies the constant threshold technique to the high-contrast circuit patterns. Experiments show that this method can properly threshold images containing both small defect patterns of approximately 20% contrast and circuit patterns of approximately 90% contrast.