產生二值化影像中物體鏈碼之平行架構

Abstract

在一張二值影像中，物體外形輪廓為最明顯的特徵，因此若能擷取到物體 輪廓，則其當對物體辨認頗有助益，同時其亦可作為求取物體的一些其他 特徵，如周長、轉角等之用，以利其他應用。然而鏈碼為描述物體輪廓最 直接的表示法，故若能快速產生鏈碼，則更能節省物體輪廓之擷取時間， 因此我們希望能用平行處理的硬體架構，使能快速產生鏈碼以即時 (real time) 擷取物體之輪廓和產生鏈碼。我們所提出的硬體架構，是一 個一維的心臟壓縮式陣列 (Systolic Array) ，可在 (n+1) 個 PE 裡藉 由規律的資料傳遞，一方面達到擷取物體輪廓的目的，另一方面則能快速 產生物體的輪廓鏈碼，並且處理一張 n × n 影像總共需花 O(3n) 的時 間，故可達到即時的要求。 In a binary image the most prominent shape feature of an object is the object contour. The contour is often used in object recognition and object feature computation such as perimeter, corner, etc. Chain code is a concise representation of the contour. A fast method for chain code generation is necessary as far as the computation time is concerned. This thesis is to design a parallel architecture to extract the object contour and generate the chain code in real time. In this thesis, we propose a parallel hardware architecture that is a one dimensional systolic array containing (n+1) PEs. After the image data in all rows are inputted in a linear fashion, the architecture extracts the contours of objects and generates the corresponding chain codes quickly. The total processing time for generating the chain codes in an (n x n) image is O(3n). This meets the real time requirement.