Process for processing images to automatically extract semantic features

摘要

In the method, correlated achromatic single-plane images are subjected to a plurality of operations for segmenting them into uniform components on the basis of different criteria. The results of the various segmentation operations for each of the processed correlated images are analyzed in order to generate cores constituted by zones presenting high certainty as to the existence of objects. The resulting cores are compared and those zones which correspond to more than some determined threshold of the various cores are conserved as indicating the presence of an object. The previously processed images are subjected to a weighted summing operation in order to obtain a first result image containing the uniform component cores. The first result image is then subjected to coloring followed by binarization in order to obtain a second result image containing the uniform components, and the individualized objects are extracted from the image.