The detection of the anomalies of the solar cells is done by testing the cells in the lab. However, this method is time consuming and expensive. The analysis of infrared solar cell images can reveal its status by classifying infrared images into anomaly and non-anomaly classes. The anomality can be due to many reasons. Therefore, it is required to not only classify image into anomaly and non-anomaly, but also, detect the anomality type. The image-based solar cell anomaly detection methods appearing in the literature used either machine learning or deep learning techniques. The main disadvantages of these methods are the lack of sufficient dataset and/or utilizing inappropriate features for classification. Machine learning requires robust feature extractor which are independent on the imaging condition. On the other hand, deep learning techniques doesn't require feature extractor, however, results depend on the implemented filters in the network i.e the network architecture. In this proposal, we deal with multi-class anomaly detection from infrared images by using better representation of the images features by using Wavelet scattering Transform (WST). The WST coefficients are stable under signal deformations and globally invariant to signal translation and rotation. Based on the simulation results, the proposed method achieved an average accuracy of 99.98%.