Use of Transputers for the Fast Determination of Shapes Corresponding to Blood Vessel Network via CCD-IR Images

概要

論文の詳細を見る
It is never a simple task to determine objects and shapes within images especially when the time requirements for this operation are severe. Object detection tasks are often facilitated by binarization of the input images such that the resulting homogenous regions define the shapes in question. Several qualitative and quantitative techniques have been developed for this purpose. These methods are usually slow, because large number of computation are requested to carry out neighborhood operations on large image data. In order to achieve a quick software and hardware system for the solution of this problem it is unavoidable to use parallel computing. In this work the use of a transputer network as the basis or the parallel image analyzer system for determining objects corresponding to blood vessels is investigated. IR filtered images of a CCD camera system connected directly to the transputer network are applied as the input of the system. We propose a parallel second derivative boundary detection based object determination technique for the extraction of medical objects. Both algorithmic and data flow parallelization are applied in order to obtain an effective parallel technique and implementing it on the actual transputer network.
一般社団法人情報処理学会の論文
1993-03-01