Camera traps are a widely used tool for analyzing and studying natural environments. Equipped with motion sensors, these cameras can capture a series of photographs when an animal passes by. However, a significant number of these images are empty, meaning no wildlife is present. Automatically detecting and discarding these empty images is a crucial task. Many existing solutions in the literature rely on supervised machine learning techniques, which require labeling numerous images. In this article, we present MORENA, an unsupervised algorithm for empty image detection, eliminating the need for costly manual labeling.