In some image processing and computer vision applications, it isdesirable to be able to extract objects of a certain width from animage. Sifting theory, which is based on operations from mathematicalmorphology provides a means to extract (`sieve out') objects up to acertain prespecified width. More often than not, the exact width of thedesired objects is not known in advance. In these cases, one would liketo have a sieve with `unsharp' meshes. Objects would pass this sieve toa certain degree. In this contribution, the author presents such anunsharp sieve, which is based on local gray level operations frommathematical morphology. It takes two dimensional binary images as inputand maps each pixel to the interval [0,1] proportional to the width ofthe object (or object part) it belongs to
展开▼
