Computer vision for automated quality inspection of colour printing products

Quality inspection of colour printing products constitutes a very important QC task in printing industry. The current practice in industry is to inspect complex colour products manually. This process is labour-intensive and the results are not reliable. The types and causes of defects on colour prints are discussed in this paper. We present a novel algorithm to automate the printing inspection process, which incorporates colour histogram-based techniques for colour image processing and a neural network for image classification. The experimental results show that this algorithm is able to detect both registration and purity defects under varying illumination conditions.