“…In top-down models, high-level signs, such as context, semantic, and knowledge information, are used to compute a saliency map. However, bottom-up models are designed based on low-level local features, such as colour, contrast, texture, intensity, motion, and orientation [21,22].…”