Browse Prior Art Database

Automatic Segmentation Method Considering the Width of the Growing Route

IP.com Disclosure Number: IPCOM000113534D
Original Publication Date: 1994-Sep-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 4 page(s) / 122K

Publishing Venue

IBM

Related People

Kimura, M: AUTHOR [+2]

Abstract

Disclosed is an algorithm for improving "the region-growing method", a well-known method for segmenting a 3-D region in a volume of image, by decreasing the tendency to leak through narrow routes. It does so by giving up any route through which a sphare of a specified radius (a controllable parameter) cannot pass. The algorithm contains two steps for checking processes: the first execludes pixels in narrow routes from the region, and the second recovers details of the region that were lost in the first step.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Automatic Segmentation Method Considering the Width of the Growing
Route

      Disclosed is an algorithm for improving "the region-growing
method", a well-known method for segmenting a 3-D region in a volume
of image, by decreasing the tendency to leak through narrow routes.
It does so by giving up any route through which a sphare of a
specified radius (a controllable parameter) cannot pass.  The
algorithm contains two steps for checking processes:  the first
execludes pixels in narrow routes from the region, and the second
recovers details of the region that were lost in the first step.

      The conventional region-growing method starts to create a
region at a seed point, specified with a pointing device by a user,
and grows the region by clustering neighboring pixels into it
according to a given condition (to be called as "the region
condition" hereafter).  When no pixels next to the region satisfy the
region condition, the growing terminates.  A typical region condition
is that the value of a pixel should lie within a given range.

      The method has the problem that it does not work well if the
region to be segmented is not completely isolated from neighboring
regions.  Narrow routes allow the region to grow larger than the user
expects.  This problem is serious in medical images, in which
tube-like objects such as blood vessels and neural fibers tend to
leak the growing region to the surrounding regions.

      The algorithm described in this article is intended to solve
this problem, and is designed to be used as an additional region
condition, together with a given one (called as Condition C).

      Fig. 1 shows a flow chart of the algorithm.  As shown in Fig.
1, the algorithm consists of two steps.

      The first, shown on the right-hand side, is for excluding
pixels on narrow routes.  It che...