I've developed an edge detection algorithm inspired by functions of the visual cortex such as divisive normalization and spatial pooling. This goal here is to sit in the class of compact methods like Canny, while having a small set of trainable parameters that adjust to data.