本論文主要利用兩種微積分的方法應用在影像特徵的求取上，第一個應用為影像邊緣檢測，利用的方法是檢查的點與周圍的梯度做不同的加權求和，並利用求和的結果作為檢測的依據。第二種方法則以仿人類視覺對突然改變的事物比較敏感為基礎，利用影像移動時最大的變化量來求得特徵點，並對位於角上的特徵點為多少角度做個估計。
我們還利用實際的影像(地圖與照片)來做邊緣檢測與特徵點檢測，並對角度取個估計值。最後對實際應用的過程中遇到的問題做個延伸的討論。最終實作的結果，角度估計的誤差達到10°以內。

關鍵字：多變數微積分、特徵檢測、電腦視覺

This paper mainly uses two kinds of calculus methods apply to find feature on image. The first application is edge detection. The method uses different weight summation between image checkpoint and surrounding points. And use the result of the sum as basis for detection. The second method is based on human vision is more sensitive to somethings sudden changes. Using maximum change when image moves to find the feature point, and estimate the angle of feature point on corner.
We also use actual images for edge detection and feature point detection, and take an estimate of the angle. Finally, the error of angle estimate is less than 10°.
Keyword：multivariate calculus, feature detector, computer vision.

[1] Richard Szeliski, Computer Vision: Algorithms and Applications, August 5, 2010.
http://szeliski.org/Book/

[2] John Canny, Finding Edge and Lines in Images, MIT Artificial Intelligence
Laboratory. https://dspace.mit.edu/handle/1721.1/6939

[3] Chris Harris & Mike Stephens , A combined corner and edge detector, United
Kingdom © The Plessey Company pic. 1988

[4] Stephen H. Friedberg, Arnold J. Insel, Lawrence E. Spence, Linear Algebra, 4th
Edition, November 21 ,2002

[5] Haldo Spontón, and Juan Cardelino, A Review of Classic Edge Detectors, Image
Processing On Line, 5 (2015), pp. 90–123. https://doi.org/10.5201/ipol.2015.35

[6] Tzu-Heng Henry Lee, Edge Detection Analysis, Graduate Institute of
Communication Engineering, National Taiwan University, 2021

[7] Javier Sánchez, Nelson Monzón, and Agustín Salgado, An Analysis and Implementation of the Harris Corner Detector, Image Processing On Line, 8
(2018), pp. 305–328. https://doi.org/10.5201/ipol.2018.229