Compact centroid distance shape descriptor based on object area normalization

Shape descriptors are more powerful to discern objects present in the images. The present work is focused on simple contour based shape descriptor using centroid distance function and it works on closed contour objects. Object area normalization is performed to obtain `N´ normalized contour points. The centroid distance feature extraction is performed on all normalized points. It forms simple 1-D feature vector of size `N´. For similarity matching correlation coefficient metric is used. This shape descriptor satisfies affine invariance property. The proposed idea is tested on MPEG-7 CE Shape-1 Part-B dataset images to validate its effectiveness. Experimental results shows that proposed compact centroid distance shape descriptor is more accurate than basic centroid distance shape descriptor and it saves space and time requirements at processing.