Fig. 2

For step 2, the image is divided into B equally-sized bins after preprocessing. Next, the feature/signature extraction for step 3 is completed by representing each bin as a \(\mathbb {R}^{3}\) vector that consist of the bin center location (x and y) and the texture feature (f) describing the pixels in the bin. The set of these B vectors is the final signature representation. The signature can also be represented as a heatmap that shows the feature value for the pixels in a region. In this example, the percentage of root pixels is used as the texture feature