Distance context based PCB film image alignment

Abstract

Purpose – This paper presents a novel printed circuit board (PCB) film image alignment method based on distance context of image components, which can be directly used for PCB film inspection. PCB film inspection plays a very important role in PCB production. Design/methodology/approach – First, image components of reference film image and inspected film image are extracted. Then, local distance context (LDC) and global distance context (GDC) are computed for each image component. Using LDC and GDC, the similarity of each pair of components between the reference film image and the inspected film image are computed, the component correspondences can be established accordingly and the parameters for aligning these two images can be eventually estimated. Findings – LDC and GDC act as the local spatial distribution descriptor and the global relative position descriptor of the current component, and they are invariant to translation, rotating and scale. Experimental results on aligning real PCB film images against various rotations and scaling transformation show that the proposed algorithm is fast and accurate and is very suitable for PCB film inspection. Research limitations/implications – The proposed algorithm is suitable for aligning those images that have some isolated connected components, such as the PCB film images. It is not suitable for general image alignment. Originality/value – We put forward to use LDC and GDC as the local descriptor and global descriptor of an image component, and designed a PCB film image alignment algorithm that can overcome the shortcomings of that image alignment algorithm that was based on local feature descriptors such as Fourier descriptor.