Key Point Detection Techniques
Key Point Detection Techniques
复制标题
关键点检测技术
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
A. Amein
中科院分区：
文献类型：
--
作者：
Abdelhameed S. Eltanany;M. Elwan;A. Amein
图像配准是找出某一区域内两幅或多幅图像之间的偏移量或错位，以确定将一幅图像中的点与另一幅图像中对应的点对齐所需的几何变换的过程。通常，配准过程的操作目标是对输入进行几何变换，从而使输入图像的几何一致，使得输入图像中的匹配像素指的是捕获区域的相同区域。因此，图像配准可以应用于变化检测、镶嵌、生成超分辨率图像等许多应用中。配准过程分为两类：(1)传统方法和(2)自动化方法。传统的方法是人工选取锚点、控制点、控制点，采用转换模型，耗时长，精度低。因此，自动检测这些点有助于恢复人工选择的性能。配准过程涉及到光照变化、亮度变化、不同传感器、噪声等诸多问题，因此，配准的应用主要依赖于采集过程中出现的误差(多时相、多视角或多模式)。特征检测作为图像配准过程的一个步骤，目的是在变化的条件下找到一组稳定的(不变的)有特色的关键点或区域。此外，探测器对视点、亮度和其他失真的变化保持健壮也是至关重要的。本论文的目的是讨论和展示常见的角点检测器，以帮助熟悉各种特征检测的应用技术。
Image registration is a process to find the offset or misalignment between two or more images for a certain area to determine the required geometrical transformation that aligns points in one image with its corresponding in the other one. Generally, the operational goal of the registration process is a geometrical transformation for the input leading to geometrically agreement for input images, so that the matched pixels in the input images refer to the same region of the captured area. So, image registration can be applied in many applications such as change detection, mosaicking, creating super-resolution images etc. Registration process is divided into two categories: (1) Traditional methods and (2) Automated methods. For the traditional methods, the anchor, control, points are selected manually and applying the transformation model leading to time consuming and low accuracy. So, automatically detection of these points helps to recover the performance of manual selection. Registration process deals with many problems such as illumination changes, intensity variations, Different sensors, noise etc. So, its applications are mainly dependent on errors (multi temporal, multi view, or multi modal) occurred during capturing process. Feature detection, as a step of image registration process, aims to find a set of stable (invariant) distinctive key points or regions under varying conditions. Also, it is critical for the detector to be robust to changes in viewpoint, brightness, and other distortions. The goal of current paper is discussion and exhibition of the common corner detectors helping to be familiar with the various applied techniques for feature detection.