Region-based feature extraction focuses on the local distribution of low level features such as colour and texture. By comparing local features in disjoint image regions, we analyse the spatial arrangement of these features. We first define a matrix of MxN disjoint regions of interest (ROIs) to sample the distribution of colour and texture information within the data set.
To extract local information the Vessel config maker device is used to create a matrix of MxN regions of interest (ROIs) that cover the area of the imaged turfgrass sample.
These ROIs are then labeled according to their position. For example “B03” represents the second row and third column. We first convert the regions to foreground blobs with Universal converter (option: binary image), and then use a second Universal converter (option: Image object list) to obtain image objects from the blobs. Then ROI labels are associated with each image object via Image object region of interest. The result of this image processing is a two-dimensional array of ROIs each with a specific ROI labels.