利用影像曲面曲率之影像表示方法

Abstract

影像之表示法在影像處理中是重要的一環。影像之表示法是將影像中 主要的特徵以一種方便處理的形式來表示。一般而言，影像之表示有許多 方式。在本篇論文中，我們提出利用影像曲面之曲率來表示影像。對於要 找的特徵點而言，對比和曲率是兩個主要的屬性。形態學對於處理幾何特 性，像大小，形狀及對比提供有效之方法。我們改變形態學所提供低通濾 波之方法提出一個能消除影像雜訊且保留影像邊緣特性的方法。同時，利 用形態學的斜率來粗略估計曲率，並使用形態學的擴張斜率及侵蝕斜率的 極值來偵測特徵點。這方法可以很容易以軟體方式或硬體方式來實現。我 們也介紹數個曲率估計之方法，並比較這些方法。 Image representation is important. Representation is to transform the imagedata into a form which contains the major features of the image and is easierto be processed. Generally, an image may be represented in many ways. In thisthesis, a scheme for the representation of an image surface by using thecurvature information is proposed. The contrast and curvature are the twomajor attributes of a feature point that we are looking for. Since "morphology" can efficiently deal with somegeometrical features, like size, shape andcontrast, a modified morphological smoothing method is used to smooth thenoise while preserving edges. The morphological gradient is used to roughlyestimate the curvature. The maxima of dilation gradient and erosion gradientare used for the detection of the feature points. This method can be easilyimplemented in both software and hardware. Several curvature estimation methods are also introduced, and the comparison of these methods is shown.