A Comprehensive Review on Anomaly Detection in Images: Challenges and Future Research Directions
Shalini Kumari, Chander Prabha · 2023
Identifying irregularities in data, or "anomalies," is essential in several fields, like medical imaging, intrusion detection (ID), fraud detection (FD), etc. A brief review of various approaches and methods presented by numerous researchers is presented in this paper. These approaches range from transformer models to image filters, autoencoders to low-rank representation, vision transformers to isolation forests, and convolutional autoencoders to deep neural networks. The primary goal of these methods is to capture the anomaly in the images. The significance of anomaly detection in numerous other applications viz. healthcare, cybersecurity, and industrial control systems are also highlighted in the paper and it also provides a brief insight into a variety of machine-learning methods to accurately detect anomalies.