A Novel Extracting Blob-like Object Method Based on Scale-Space
Wenbing Chen, Xia Wang, Qizhou Li, Yunjie Chen · 2011
This paper presents a novel method, which can be used to extract blob-like object from a blob-like image. The method firstly uses an interest point detecting algorithm with automation scale selection to detect interest points and their scales. Secondly, centered at each interest point a local rectangle region can be constructed with two scales of the point in two different directions. Since such a region contains a single object, the set of these local regions can be regarded as an approximate segmentation for the original image. Further more, we can use a clustering method to extract a blob object from each local region. Experimental results show that our method can efficiently extract single object.