The scipy.misc.imresize function is a little strange for me. First, this is what happens when I specify the image of the 2D image that you provided to call scipy.misc.imresize this image with a scale of 1.0. Ideally, it should give you the same image, but we get this (in IPython):
In [35]: from scipy.misc import imresize In [36]: import numpy as np In [37]: dtest = np.array(([1,2,3],[4,5,6],[7,8,9])) In [38]: out = imresize(dtest, 1.0) In [39]: out Out[39]: array([[ 0, 32, 64], [ 96, 127, 159], [191, 223, 255]], dtype=uint8)
It not only changes the type of output to uint8 , but also scales the values. Firstly, it looks like the maximum value of the image is 255, and the minimum value is 0. MATLAB imresize does not do this and imresize image as we expect:
>> dtest = [1,2,3;4,5,6;7,8,9]; >> out = imresize(dtest, 1) out = 1 2 3 4 5 6 7 8 9
However, you need to know that MATLAB performs resizing with anti-aliasing enabled by default . I'm not sure scipy.misc.resize here, but I bet that anti-aliasing is not turned on.
Edit - November 23, 2016
As Eric noted in his comments below, if you pre-draw the image on the desired type, you will get the expected results:
In [10]: dtest = np.array([[1,2,3],[4,5,6],[7,8,9]], dtype=np.uint8) In [11]: out = imresize(dtest, 1.0) In [12]: out Out[12]: array([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=uint8)
We see that the image does not scale to the range [0,255] . To finally get where you need to go, we need to get a floating-point representation of the image. scipy.misc.imresize has an additional flag named 'mode' , and you can specify it as 'F' to ensure that the output is a floating point.
In [14]: scale = 1.4 In [15]: out = imresize(dtest, 1/scale, mode='F') In [16]: out Out[16]: array([[ 2.5 , 3.75], [ 6.25, 7.5 ]], dtype=float32)
As you will see below, the results that you see with scipy.misc.resize do not match what you see in MATLAB.
For best results, do not specify the scale - specify the target output size for reproducing the results. So 1/scale in your case is close to a 2 x 2 output, and here is what you would do in MATLAB:
>> dtest = [1,2,3;4,5,6;7,8,9]; >> out = imresize(dtest, [2,2], 'bilinear', 'AntiAliasing', false) out = 2.0000 3.5000 6.5000 8.0000
You can see that some of the values in the matrix do not align with scipy.misc.resize . To match what you see in MATLAB. Closest to what you want is OpenCV resize or scikit-image resize . Both of them have no smoothing. If you want both Python and MATLAB to match, use the bilinear interpolation method. imresize in MATLAB uses bicubic interpolation by default, and I know that MATLAB uses custom kernels for this, and therefore it will be much more difficult to compare their results if you use bicubic interpolation between these methods. See this post for more informative results:
MATLAB vs C ++ vs OpenCV - imresize
With Python OpenCV:
In [93]: import numpy as np In [94]: import cv2 In [95]: dtest = np.array(([1,2,3],[4,5,6],[7,8,9]), dtype='float') In [96]: out = cv2.resize(dtest, (2,2)) In [97]: out Out[97]: array([[ 2. , 3.5], [ 6.5, 8. ]])
With scikit-image:
In [100]: from skimage.transform import resize In [101]: dtest = np.array(([1,2,3],[4,5,6],[7,8,9]), dtype='uint8') In [102]: out = resize(dtest, (2,2), order=1, preserve_range=True) In [103]: out Out[103]: array([[ 2. , 3.5], [ 6.5, 8. ]])
It is interesting to note that MATLAB, OpenCV, and the scikit image, when defining a floating point scale, act differently with each other. I did some experiments and, indicating the size of the floating point, I could not find the results. In addition, scikit-image does not support the adoption of a scale factor, which is more a basis for explicitly indicating the size of the output, rather than the scale.