论文标题
IM：用于计算图像矩和力矩不变的R包装
IM: An R-Package for Computation of Image Moments and Moment Invariants
论文作者
论文摘要
力矩不变是图像分类的良好和有效的形状描述符。在本报告中，我们介绍了一个名为IM的R语言的软件包，该软件包实现了图像矩的计算，并允许从面向对象的框架内的矩重建图像。可以使用IM库来计算几种类型的时刻，包括离散和连续的Chebyshev，Gegenbauer，Legendre，Krawtchouk，Dual Hahn，广义伪Zernike，Fourier-Mellin和Radial Harmonic Fourier Moments。另外，可以使用两种不同类型的多项式的组合来计算自定义双变量矩的类型。像素坐标的极性转化方法用于在矩形上正交的矩中旋转近似不变性。本报告中讨论了用于计算矩的不同类型的多项式，以及重建和运行时间的比较。提供了使用图像矩的图像分类的示例。
Moment invariants are well-established and effective shape descriptors for image classification. In this report, we introduce a package for R-language, named IM, that implements the calculation of moments for images and allows the reconstruction of images from moments within an object-oriented framework. Several types of moments may be computed using the IM library, including discrete and continuous Chebyshev, Gegenbauer, Legendre, Krawtchouk, dual Hahn, generalized pseudo-Zernike, Fourier-Mellin, and radial harmonic Fourier moments. In addition, custom bivariate types of moments can be calculated using combinations of two different types of polynomials. A method of polar transformation of pixel coordinates is used to provide an approximate invariance to rotation for moments that are orthogonal over a rectangle. The different types of polynomials used to calculate moments are discussed in this report, as well as comparisons of reconstruction and running time. Examples of image classification using image moments are provided.
