Development of control charts to monitor image data using the contourlet transform method

Abstract

In recent years, researchers and practitioners have been exploring new methods for quality control, including image processing. The effective use of high-volume image data can significantly improve the monitoring of production and service systems in terms of speed, accuracy, and cost. Adopting an image-based approach is better than relying on operator-based solutions, and it offers new perspectives for process monitoring. Image processing can involve extracting features to identify, classify, detect, and cluster. Although there are several transformations to extract features from images, the Fourier method cannot consider the concurrency of frequency and time data, and the wavelet method only considers two specific directions. In multidimensional transforms, the optimal method can provide more information using fewer coefficients. The contourlet transform has advantages such as multiresolution, localization, critical sampling, directionality, and anisotropy. This research investigates the advantages of applying the contourlet transform to images and using data in a generalized likelihood ratio control chart. The results show that this method is more accurate than others because it can examine various directions in images. The proposed methodology algorithm is also presented in this study.