图像的仿射变换

准备仿射变换 (Affine2D ） data transform 一个图像的形状允许操纵图像的形状和方向。这是一个例子的概念 transform chaining .

输出图像的边界应与黄色虚线矩形匹配。

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.transforms as mtransforms


def get_image():
    delta = 0.25
    x = y = np.arange(-3.0, 3.0, delta)
    X, Y = np.meshgrid(x, y)
    Z1 = np.exp(-X**2 - Y**2)
    Z2 = np.exp(-(X - 1)**2 - (Y - 1)**2)
    Z = (Z1 - Z2)
    return Z


def do_plot(ax, Z, transform):
    im = ax.imshow(Z, interpolation='none',
                   origin='lower',
                   extent=[-2, 4, -3, 2], clip_on=True)

    trans_data = transform + ax.transData
    im.set_transform(trans_data)

    # display intended extent of the image
    x1, x2, y1, y2 = im.get_extent()
    ax.plot([x1, x2, x2, x1, x1], [y1, y1, y2, y2, y1], "y--",
            transform=trans_data)
    ax.set_xlim(-5, 5)
    ax.set_ylim(-4, 4)


# prepare image and figure
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2)
Z = get_image()

# image rotation
do_plot(ax1, Z, mtransforms.Affine2D().rotate_deg(30))

# image skew
do_plot(ax2, Z, mtransforms.Affine2D().skew_deg(30, 15))

# scale and reflection
do_plot(ax3, Z, mtransforms.Affine2D().scale(-1, .5))

# everything and a translation
do_plot(ax4, Z, mtransforms.Affine2D().
        rotate_deg(30).skew_deg(30, 15).scale(-1, .5).translate(.5, -1))

plt.show()

---

工具书类

本例中显示了以下函数、方法和类的使用：

import matplotlib
matplotlib.axes.Axes.imshow
matplotlib.pyplot.imshow
matplotlib.transforms.Affine2D