基于鞍点搜索的粘连对象图像分割方法.pdfVIP

35 10 2009 5 Vol.35 No.10 Computer Engineering May 2009 ·· 2009 A TP391 ( 315211) Hessian () Hessian Hessian Image Segmentation Method of Touching Objects Based on Saddle-point Search WANG Yi-cong, CHEN Ken (College of Information Science and Engineering, Ningbo University, Ningbo 315211) AbstractAiming at the problem of segmenting the touching anomalous objects in an image, this paper proposes a method based on saddle-point detection using Hessian matrix. The image is converted to its 3D topographic counterpart by sequential erosion, followed by searching the saddle-point utilizing Hessian matrix determination approach. The segmentation is performed at the identified saddle-point by tracking the maximal gradient path. Compared with other conventional segmentation algorithms, this method is more practical for implementation, and shows its superiority in execution speed. Key wordssaddle-point; Hessian matrix; eigenvalue 1 Taylor n (n=2) f x0 f (x 0 +∆x ) f (x 0 ) =+∆x T∇f (x 0 ) +∆x TH (x 0 )∆x (1) ∇f H (x ) 0 Hessian [1-2] (watershed ⎡f f ⎤ H ⎢ xx xy ⎥ (2) segmentation)(geodesic reconstruction) ⎢f f ⎥