GeezerVison is showing github.com — [Bug]: I give an RGB image to imsave but I don't have the right color map! · Issue #29183 · matplotlib/matplotlib
[Bug]: I give an RGB image to imsave but I don't have the right color map! #29183DescriptionBug summary
When I give an RGB image as a list of list of list to imsave, I don't have the right colormap and there is no rgb colormap!
Code for reproduction
import matplotlib.pyplot as plt
I = plt.imread("notre-dame.jpg")
m,n,p = I.shape #m=largeur, n=hauteur, p=uplet
J=[[[0 for d in range(p)] for e in range(n)] for f in range(m)]
K=[[[0 for d in range(p)] for e in range(n)] for f in range(m)]
L=[[[0 for d in range(p)] for e in range(n)] for f in range(m)]
for i in range(m) :
for j in range(n) :
J[i][j]=[int(I[i][j][0]),0,0]
K[i][j]=[0,int(I[i][j][1]),0]
L[i][j]=[0,0,int(I[i][j][2])]
fig, ax=plt.subplots(4,figsize=(28,10))
ax[0].imshow(I)
ax[1].imshow(J)
ax[2].imshow(K)
ax[3].imshow(L)
ax[0].set_axis_off()
ax[1].set_axis_off()
ax[2].set_axis_off()
ax[3].set_axis_off()
ax[0].set_title('Image complete')
ax[1].set_title('Composante rouge')
ax[2].set_title('Composante bleue')
ax[3].set_title('Composante verte')
fig.suptitle("Image de Notre-Dame", x=0.51,y=.93)
plt.imsave("Notre-Dame_rouge.png",J) #The problem is here!
plt.savefig("Notre-Dame_composante_ex1.png",dpi=750,bbox_inches="tight")
plt.savefig("Notre-Dame_composante_ex1.pdf",dpi=750,bbox_inches="tight")
plt.show()Actual outcome
Expected outcome
The red one in a png image!
Additional information
The original image
Operating system
Win 11
Matplotlib Version
3.9.2
Matplotlib Backend
module://matplotlib_inline.backend_inline
Python version
3.11.10
Jupyter version
No response
Installation
conda Activitytacaswell commented on Nov 25, 2024 The source of the problem is that in figimage we only check for RGB if the input is already a numpy array. Is there a reason you need to work with list-of-list-of-list? Using numpy (which I is already a numpy array) the code becomes simpler (no nested loop), faster (numpy does array work at or near c-speed), and avoids the bug:
import matplotlib.pyplot as plt
import numpy as np
I = plt.imread("notre-dame.jpg")
J = np.zeros_like(I)
# RGB version of the image with only one plane non-zero
J[..., 0] = I[..., 0]
K = np.zeros_like(I)
K[..., 0] = I[..., 1]
L = np.zeros_like(I)
L[..., 0] = I[..., 2]
# share the axis so that panning / zooming on one pans/zooms on all
fig, axs = plt.subplots(4, figsize=(28, 10), sharex='all', sharey='all')
# loop over axes, data, and title strings
for ax, data, title in zip(
axs,
(I, J, K, L),
("Image complete", "Composante rouge", "Composante bleue", "Composante verte"),
):
ax.imshow(data)
ax.set_title(title)
fig.suptitle("Image de Notre-Dame", x=0.51, y=0.93)
fig.savefig("Notre-Dame_composante_ex1.png", dpi=750, bbox_inches="tight")
fig.savefig("Notre-Dame_composante_ex1.pdf", dpi=750, bbox_inches="tight")
plt.imsave("Notre-Dame_rouge.png", J) # The problem is here!
plt.show()As a short-term fix you can also do np.array(J) before passing to fig.imsave which will go through the correct code path. github-actions commented on Nov 25, 2024 Good first issue - notes for new contributors
This issue is suited to new contributors because it does not require understanding of the
Matplotlib internals. To get started, please see our contributing
guide.
We do not assign issues. Check the Development section in the sidebar for linked pull
requests (PRs). If there are none, feel free to start working on it. If there is an open PR, please
collaborate on the work by reviewing it rather than duplicating it in a competing PR.
If something is unclear, please reach out on any of our communication
channels. tacaswell commented on Nov 25, 2024 This is a good first issue as it should require no API choices and we have a good reference for the expected behavior already in the code base.
The issue is that when passing a list-of-list-of-list to plt.imsave and saving to a format that pillow and write the check that sees if it is an RGB image (which should not be color mapped) or a 1-channel luminescence image (which should) is avoided and we always color map.
The steps are:
- look at
Axes.imshow to see how we do the input normalization there
- do the same thing is
matplotlib.image.imsave
- add a test (round-trip with
imread and a small synthetic RGB image)
Frostbyte72 commented on Nov 25, 2024 Hi I am a new contributor to opensource and believe I can fix this could I be assigned to this issue? tacaswell commented on Nov 25, 2024 zef126 commented on Nov 25, 2024 Thank you so much for your help...it is very useful, I will try to use arrays!
The point is that I am in preparation class in France and parts of the code are strictly imposed by the teacher! zef126 commented on Nov 25, 2024 It is the first repo where i fond kindness and attention, it is really appreciated! zef126 commented on Nov 25, 2024 Now it's just a code issue...I don't understand why I get that:
Instead of that:
with your code @tacaswell !
And by the way how can i save my lists in rgb files ...if it's possible?
I tried to convert them in numpy.arrays like that:
plt.imsave("Notre-Dame_rouge.png", np.array(J))
but it returned that:
ValueError: Image RGB array must be uint8 or floating point; found int64
Thank you again for your help! QuLogic commented on Nov 26, 2024
J[..., 0] = I[..., 0]
K = np.zeros_like(I)
K[..., 0] = I[..., 1]
L = np.zeros_like(I)
L[..., 0] = I[..., 2]
K and L should use 1 and 2, respectively.
tacaswell commented on Nov 26, 2024 🤦🏻
as noteed, should be
J[..., 0] = I[..., 0]
K = np.zeros_like(I)
K[..., 1] = I[..., 1]
L = np.zeros_like(I)
L[..., 2] = I[..., 2] and np.asarray(J, dtype=np.int8) zef126 commented on Nov 26, 2024 Thank you very much!
I just got this for saving the list of list of list:
plt.imsave("Notre-Dame_rouge.png", np.asarray(J, dtype=np.int8))
OverflowError: Python integer 128 out of bounds for int8aragan22 commented on Feb 18, 2025
Below is the full code for the script, including the fix for the imsave function to handle nested lists properly. I included headers to describe exactly what each code block function as and how it was changed.
@aragan22 This is not adding meaningfully to the issue. You are only rewriting the original restating the original minimal example with comments - but the example and issue is well understood. Nobody with the capability of fixing the issue needs these comments. Please think carefully to not post repetitive or unneeded stuff.
Apologies, I did not mean to comment that portion. That was simply for myself as I tried to understand the issue and work around it in my own way. Thank you for pointing this out, apologies again for the confusion. Dronten commented on Mar 22, 2025 is this issue still up and can i work in this ? rcomer commented on Mar 22, 2025 Aftabby commented on Apr 8, 2025 Has the issue been solved? Else I'd like to take over. timhoffm commented on Apr 9, 2025 guillermodotn commented on Apr 16, 2025 I have opened PR #29931 for this issue, based on the latest code review in #29203. jk2500 commented on May 6, 2025 Hey there, I tried running the code in the snippet and it shows the expected output. Does this mean the issue can be closed? timhoffm commented on May 7, 2025
MetadataMetadataDevelopmentNo branches or pull requestsIssue actions
|
