多處理器平行處理架構(Parallel processing)之用途廣泛,包括核子試爆模擬分析、車體撞擊測試、以及高速影像處理運算等應用。而影像多值門檻分隔乃是一基本且重要之影像處理法則,然目前文獻大都探討單一處理器之應用,多處理器平行處理之探討目前仍非常缺乏。 本論文的貢獻在於改進Otsu[1]二值及多值影像門檻分隔演算法之效率且使其適用於平行處理運算,再搭配個人電腦叢集式系統(PC Clustering System)與高速區域網路為硬體架構,實踐即時的影像平行處理。實驗結果在二值及多值門檻分隔處理時都可得到預期的效率,因此證明使用本文所提之叢集式系統與所推導之影像多值門檻分隔平行處理演算法非常適合做影像平行處理。
Parallel processing system wide range of application includes the nuclear test emulation analysis, bodywork to bump the shot test and super-speed image processing etc. The multi-level image thresholding is a basic and count for image processing rule, however the cultural heritage aims at the one processor and developments mostly, the related cultural heritage of the image parallel processing is still very limited. This paper contribution in improving the Otsu [1] two-level and multi-level image thresholding of Otsu’s technique and achieved a faster parallel processing version of Otsu’s, then match a gather type in PC Cluster System and super-speed LAN as the hardware to configure, practicing the instant image parallel processing, experiment result at Between-Class and multi-level image thresholding off the processing can get the anticipant efficiency, so the certificate uses Cluster System that this text lift and multi-level image thresholding parallel processing algorithm to suit to do the image parallel processing.