Image module¶
The image module implements the Image class. Image class provides conversion primitives and transformation functions to work on your data.
Conversion¶
Let us create an empty image and export it
import numpy as np
import mpstool.img as mpsimg
# creates an image from a numpy array
image = mpsimg.Image.fromArray(np.zeros((10,10)))
# creates an image from a txt dump of numpy
# The shape of the input should be provided here
image = mpsimg.Image.fromTxt("2D.txt",(550,500))
# creates an image from a gslib file
image = mpsimg.Image.fromGslib("2D.gslib")
# creates an image from a vox file
image = mpsimg.Image.fromVox("2D.vox")
We can export it to various file formats:
image.exportAsPng("black.png")
image.exportAsGslib("black.gslib")
image.exportAsVox("black.vox")
image.exportAsTxt("black.txt")
Supported formats are txt (numpy dump), gslib, png, vox, vtk, pgm, ppm
The 3D example is analogous. Be careful however, that some file formats like png and txt will not work:
image = mpsimg.Image.fromArray(np.zeros((10,10,10)))
image.exportAsGslib("black.gslib")
image.exportAsVox("black.vox")
Image class support multiple variables. This allow to define several channels, and output images in colors:
data = np.array([np.zeros((10,10,1)) for i in range(3)]) #define three channels for a 2D picture of size 10x10
image = mpsimg.fromArray(data)
image.rename_variable("V0", "R")
image.rename_variable("V1","G")
image.rename_variable("V2","B")
image.exportAsPng("black.png",colored=True)
image.exportAsGslib("black.gslib") # will output the three variables side by side
Image modification¶
One can add, remove of modify the variables in an image:
data = np.zeros((10,10,10))
image.add_variable(data,"new_var")
image.rename_variable("new_var","my_var_name")
image.remove_variable("my_var_name")
One can also perform several operations, like flipping or permuting axis:
image.flipx()
image.permxy()
Normalizing¶
Images contain either integer values from 0 to 255 (denormalized) or float numbers between -1 and 1 (normalized). Methods .normalize() and .unnormalize() allow you to switch between the two.
The normalize parameter is set to False by default in the importation functions.
Transformation functions¶
Categorize:
image.categorize(3)
# will discretize the image to have only 3 different values
# The discretization algorithm is a k-means where k is the number of categories
# Each category will then take the color of the majority of pixels within it
Threshold:
image.threshold(thresholds=[100,200], values=[10,50,150])
# this will transform the image the following way:
# values < 100 will be sent to 10
# values >=100 but <200 will be sent to 50
# values >=200 will be sent to 150
Saturation of the white:
image.saturate(t=5)
# Values below t will be sent to 0
You can also apply your custom transformation on the values of the pixels:
f = lambda x : (x+15)%255
image.apply_fun(f)
Sampling, cutting, tilling¶
The library also provides functions to extract samples from an image:
image = mpsimg.Image.fromPng("2D.png")
sample_shape = (20,20)
sample = image.get_sample(sample_shape)
sample.exportAsPng("sample.png")
The cutting function is useful on 3D images. It outputs every cuts of the data along a given axis. The n parameters tells how many random cuts should be taken. If n equals -1, all cuts are taken
image = mpsimg.Image.fromTxt("3D.txt",(100,90,80))
image.exportCuts("cuts_folder",axis=0,n=-1)
# axis = 0 <-> x ; 1 <-> y ; 2 <-> z
The tilling function takes a list of images and tile them together into a single one:
image1 = Image.fromGslib("2D.gslib")
image2 = Image.fromGslib("2D.gslib")
image3 = Image.fromGslib("2D.gslib")
image4 = Image.fromGslib("2D.gslib")
image_list = [image1, image2, image3, image4]
tiled_horizontal = Image.tile_images(image_list, 'h')
tiled_vertical = Image.tile_images(image_list, 'v')
tiled_square = Image.tile_images(image_list, 's')