Fuzzy shape Classification exploiting Geometrical and Moments Descriptors

In the era of data intensive management and discovery, the volume of images repositories requires effective means for mining and classifying digital image collections. Recent studies have evidenced great interest in image processing by "mining" visual information for objects recognition and retrieval. Particularly, image disambiguation based on the shape produces better results than traditional features such as color or texture. On the other hand, the classification of objects extracted from images appears more intuitively formulated as a shape classification task. This work introduces an approach for 2D shapes classification, based on the combined use of geometrical and moments features extracted by a given collection of images. It achieves a shape based classification exploiting fuzzy clustering techniques, which enable also a query-by-image.