線圖形中封閉特性的研究

On the Study of Closure Property for Line Patterns

林再得(Tsay-Der Lin)

指導教授 ： 陳永盛

元智大學/工程學院/電機與資訊工程研究所/碩士(1993年)

摘要

封閉性對於線圖形視知覺的認知提供一個很重要的特性，在於它直覺反映
出圖形群組織的基本性質，為進一步探討封閉特性，我們提供二個方法以
電腦視覺的立場去從事雜訊處理之研究。第一個方法又稱為圖形擴展法，
其特點在於以邊界性質為基準去模擬封閉特性。第二個方法又稱為視野法
，其特點在於以神經心理學中所謂的刺激側擴散為基準去模擬封閉特性。
將以上兩種方法結合成一個新模型，再用視野法去模擬人類視知覺對遠近
現象的認知。最後，我們將以上所提供的兩種方法分別與鬆弛法或形態法
做比較。經由實驗結果的証明可知，我們所提出的方法不但具有可行性，
且亦能獲得良好的結果。

關鍵字

雜訊平滑化 ； 封閉性 ； 大小恆常性

並列摘要

Closure property can provide a key to our understanding of the
visual perception of line pattern. It is a fundamental
property of pattern grouping in human visual perception as
advocated by the Gestalt school of psychology. The closure
possesses a pro- perty to organize the world by filling in the
gaps in the stimulation so that we perceive a whole Object and
disjointed parts. To explain this property, two hypothetical
models from the standpoint of computer engineering are
presented in the thesis. The one is called disk expansion
based on boundary pro- perty to simulate the closure property.
The other is called visual field method based on lateral spread
of excitation (gaussian function) in neurophysilology, and then
two proposed methods are combined in new model. Attempts also
been made to compare the performance of the proposed algorithm
with other existing algorithm (e.g. morphological algorithm and
relaxation method). The experimental results indicate that the
proposed algorithms yield the better smoothing results.
Finally, the visual field method is used to simulate the
phenomenon of far view and near view is feasible and reasonable.

並列關鍵字

noise smooth ； closure ； size constancy