I've a numpy array of shape (960, 2652), I want to change its size to (1000, 1600) using linear/cubic interpolation.
>>> print(arr.shape)
(960, 2652)
I've check this and this answer, which recommends to use scipy.interpolate.interp2d, but what should I provide as x and y?
from scipy import interpolate
f = interpolate.interp2d(x, y, arr, kind='cubic')
The datapoints on the old domain, i.e.
import numpy as np
from scipy import interpolate
from scipy import misc
import matplotlib.pyplot as plt
arr = misc.face(gray=True)
x = np.linspace(0, 1, arr.shape[0])
y = np.linspace(0, 1, arr.shape[1])
f = interpolate.interp2d(y, x, arr, kind='cubic')
x2 = np.linspace(0, 1, 1000)
y2 = np.linspace(0, 1, 1600)
arr2 = f(y2, x2)
arr.shape # (768, 1024)
arr2.shape # (1000, 1600)
plt.figure()
plt.imshow(arr)
plt.figure()
plt.imshow(arr2)
skimage.transform.resize is a very convenient way to do this:
import numpy as np
from skimage.transform import resize
import matplotlib.pyplot as plt
from scipy import misc
arr = misc.face(gray=True)
dim1, dim2 = 1000, 1600
arr2= resize(arr,(dim1,dim2),order=3) #order = 3 for cubic spline
print(arr2.shape)
plt.figure()
plt.imshow(arr)
plt.figure()
plt.imshow(arr2)
for new version of spicy (>1.10)
import numpy as np
import spipy as sp
def inter_image(image, size_out):
x_size = image.shape[0]
y_size = image.shape[1]
x = np.linspace(0, 1, x_size)
y = np.linspace(0, 1, y_size)
f = sp.interpolate.RegularGridInterpolator((y, x), image)
x = np.linspace(0, 1, size_out[0])
y = np.linspace(0, 1, size_out[1])
xg, yg = np.meshgrid(x, y)
return f((yg, xg))
