부산대학교 도서관

In this paper, we delve into a thorough analysis of various metrics aimed at bolstering anomaly detection capabilities. We assess the efficacies of various metrics, including MSE, PSNR, SSIM, NCC and Euclidean Distance in capturing anomalies in an image. Our research endeavor encompasses experimentation and analysis to evaluate the performance of each metric. Through testing on various image data sets, we find the strengths and limitations of individual metric in anomaly detection scenarios. By discerning the most optimal metric, we contribute to the advancement of anomaly detection methodologies, facilitating improved image analysis and interpretation across diverse domains. The outcome of our study holds significant implications for various industries, supporting the development of robust and accurate anomaly detection systems.