Converting height data from CSV file to IMG

Question

I have some height data from a microscope. My question is: How can I convert CSV files to an IMG output and scale the color correctly? The data I have are CSV files like:

90,40,50,30,20
11, 2, 7,30,20
90,40,50,30,50
16,40,50, 4, 6
90, 4,50,30,20
...

Where the columns are the x coordinates, the rows are y coordinates, and the entries are the z coordinates.

Also it should be over a fixed scale. That means that the generated image should not be generated over the highest and lowest value, so that multiple files are comparable.

What I did so far is:

IMG = genfromtxt('RohCSV/{}.csv'.format(filename), delimiter=',')

That kind of works, but I have the scale problem. The results are not comparable.

I hope it is kind of understandable.

Huh? If each row contains an X, a Y and a Z value, why are there 5 fields? Oh, no, you say the columns are the x-coordinates and the rows are the y-coordinates but I don't get that - if the first column is x and that's 90 that means x is 90 but the rows are the y-coordinates so 90 is the row and the column. No, that must be wrong too... what does the first row mean in plain English please? — Mark Setchell, Jan 12 '20 at 19:36
@MarkSetchell That's already the tabular (image) data. Otherwise `genfromtxt` shouldn't work that easily as presented in the question. The comment indeed is misleading, but it only tells us the typical NumPy (image) array indexing (rows = y coordinate, columns = x coordinate). I assume, the only (remaining) question is the "scale", i.e. a min/max normalization from my point of view. — HansHirse, Jan 13 '20 at 05:43

HansHirse · Accepted Answer · 2020-01-14T07:12:31.277

It depends on how you compare outputs. When using Matplotlib for example, you should get a min/max normalized output per default. Let's see the following example:

from matplotlib import pyplot as plt
import numpy as np

plt.figure(1, figsize=(12, 5))

# Original data/image
IMG = np.genfromtxt('test.csv', delimiter=',')
plt.subplot(1, 2, 1), plt.imshow(IMG), plt.colorbar()

# Min/max normalized data/image
IMG = IMG - np.min(IMG)
IMG = IMG / np.max(IMG)
plt.subplot(1, 2, 2), plt.imshow(IMG), plt.colorbar()

plt.tight_layout()
plt.show()

That'd be the output:

As you can see, the default output of Matplotlib (left) as well as the output of the explicitly min/max normalized data/image (right) are equal.

Apart from that, the presented min/max normalization should be, what you're looking for.

Hope that helps!

-----------------------
System information
-----------------------
Python:      3.8.1
Matplotlib:  3.2.0rc1
NumPy:       1.18.1
-----------------------

EDIT: After clarification in the comment, I guess using the vmin and vmax parameters in Matplotlib's imshow might be the simplest way to go:

from matplotlib import pyplot as plt
import numpy as np


def plot_scaled(data, min_scale, max_scale):
    plt.imshow(data, vmin=min_scale, vmax=max_scale), plt.colorbar()


# Original data/image
IMG = np.genfromtxt('test.csv', delimiter=',')

# Artificial data
IMG_large = IMG + 10
IMG_even_larger = IMG + 100

plt.figure(1, figsize=(15, 4))
for i, img in enumerate([IMG, IMG_large, IMG_even_larger]):
    plt.subplot(1, 3, i+1), plot_scaled(img, 0, 200)
plt.tight_layout()
plt.show()

Value 100 has the same color in all three plots.

Hi HansHirse, thank you for your answer. Your code works very well! The only problem I have left is, that the `np.min(IMG)` and `np.max(IMG)` are different from image to image. My goal is, that for example the height 5000000 is displayed always in the same color in all images. I think I need a fixed scale from height to color but I don't know how to accomplish that. For now my output is: https://imgur.com/a/foa3JXg Thanks a lot for your help so far! Best regards Max — Max, Jan 13 '20 at 17:01
@Max I updated my answer. For visualization purposes with Matplotlib for example, the easiest way to achieve that is to use the `vmin` and `vmax` parameters. — HansHirse, Jan 14 '20 at 07:17

Converting height data from CSV file to IMG

1 Answers1