Method proposal for blob separation in segmented images
Gabriel-Mihail Danciu · 2017
Image segmentation has been and still is a challenge in computer vision. Regardless of the methods used to perform this task, almost always the result is the same: a set of areas representing locations of various items found in the image. If we wish to classify the areas into two types of data, the problem becomes separating foreground from background. Our proposed method starts with this supposition: that the image is segmented and as a result we obtain a mask where the white pixels represent foreground and the black ones are the background. Another supposition we make, is that the items have almost the same size and shape. The goal is to obtain the locations of all the items in the image. This algorithm has numerous applications such as item localization, measurement, validation, etc.