Problem with Colorbars, where intensity value 0 refers to no colour (e.g., transparent)

[FloskelSalomon]

Dear Robert,

recently I encountered a error message using the stackview.imshow() function.

What I am trying to do is adding a scalebar to the visualisations of my images.

My approach is as follows, I have an image array and copy an empty array only containing "0" in the same sizes as the image array. Afterwards I take the array only containing "0" and turn the values into "1" where I want to have the scale bar. Appropriately calculated using the microns_per_pixel from the metadata of my images. The result is an array containing only "0" and "1".

In a function the code looks like this:

def scale_bar_image(image, 
                    micron_length, 
                    scale_length,
                    scale_color="white",
                    thickness=4,
                    bottom_margin=0.93,
                    left_margin=0.07):                   
    
    """
    Generates an image with a scale bar at the bottom left side
    
    Parameters:
    - image (numpy.ndarray): The input image (2D numpy array) which will be used to generate the scale bar image.
    - micron_length (float): The physical length (in microns) represented by the scale bar.
    - scale_length (float): The desired length of the scale bar in the generated image.
    - scale_color (str, optional): Color of the scale bar. Default is "white".
    - thickness (int, optional): Thickness of the scale bar. Default is 4 pixels.
    - bottom_margin (float, optional): The position of the scale bar from the bottom as a fraction of the image height. Default is 0.93.
    - left_margin (float, optional): The position of the scale bar from the left as a fraction of the image width. Default is 0.07.
    
    Returns:
    - scale_bar (numpy.ndarray): The image with the overlaid scale bar.
    - colormap (matplotlib.colors.ListedColormap): Colormap used for the scale bar.

    Raises:
    - ValueError: If the input image does not have 2 dimensions.
    """
    
    if len(image.shape) != 2:
        raise ValueError("Out of index. 2-dimensional images are required.")
    
    from matplotlib.colors import ListedColormap
    import numpy as np

    # Create a colormap for drawing the scale bar
    # Intensity value 0 is "none" for transparency
    colormap = ListedColormap(["none", scale_color])

    # Create an empty copy of the image where the scale bar should be drawn
    scale_bar = np.copy(image)
    scale_bar[:] = 0

    # Calculate the required pixels that need to be drawn for the scale bar
    length_pixels = int(scale_length / micron_length)

    # Define a margin from the bottom and the left side
    margin_bottom = int(scale_bar.shape[0] * bottom_margin)
    margin_left = int(scale_bar.shape[1] * left_margin)

    # Draw the scale bar 
    # E.g., at the position of the scale bar everything will be turned into "1"
    # Afterwards the scale_bar image can be overlayed with the colormap above
    scale_bar[margin_bottom:margin_bottom + thickness, 
              margin_left:margin_left + length_pixels] = 1
    
    return scale_bar, colormap

Finally I define a colormap using colormap = plt.color.ListedColormap(["none", "white"]). With that approach the value "0" is transparent. With that I was easily able to overlay a scalebar when the imshow function was still implemented in pyclesperanto (check code here).

However, when I am trying to do the same approach in stackview.imshow. I receive the following error message:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[7], line 15
      7 sv.imshow(np.mean(image_c1[0]["wDataCh0"][1:, ...], axis=2),
      8           plot=axs,
      9           colormap="pure_green",
     10           min_display_intensity=p1,
     11           max_display_intensity=p99,
     12           continue_drawing=True)
     14 # Add scale bar on top
---> 15 sv.imshow(scalebar,
     16           plot=axs,
     17           colormap=cmap)
     19 #plt.savefig(plots + 'M1_RR_RGC3_chirp_C1.png', dpi=300, bbox_inches="tight")

File ~/.local/lib/python3.10/site-packages/stackview/_imshow.py:82, in imshow(image, title, labels, min_display_intensity, max_display_intensity, plot, colorbar, colormap, alpha, continue_drawing, axes, cmap, vmin, vmax)
     79 if colormap is None:
     80     colormap = "Greys_r"
---> 82 if colormap.startswith("pure_"):
     83     colormap = create_colormap(colormap)
     85 if labels:

AttributeError: 'ListedColormap' object has no attribute 'startswith'

Do you have any quick solutions to the problem? Also, do you think you will implement a scale bar argument to stackview's imshow function, similar as it can be done in microfilm?

Many thanks in advance,
Florian

[haesleinhuepf]

git-bob comment with a potential solution for fixing this bug in stackview, the current repository.

[github-actions]

^{This message was generated by git-bob (version: 0.22.0, model: gpt-4o-2024-08-06, log), an experimental AI-based assistant. It can make mistakes and has limitations. Check its messages carefully.}

Hello Florian,

The issue seems to be with the type of colormap expected by stackview.imshow(). It expects a string when checking colormap.startswith("pure_"), but you are passing a ListedColormap object directly.

To resolve this, you can convert your ListedColormap back to a format that stackview.imshow() expects or bypass that check. Here's a quick snippet on how you can try bypassing it:

def modified_imshow(colormap, **kwargs):
    # Wrap the colormap check to avoid using the startswith method for non-string types
    if isinstance(colormap, str) and colormap.startswith("pure_"):
        colormap = create_colormap(colormap)
    # Use the rest of the logic from stackview.imshow with provided kwargs
    stackview.imshow(colormap=colormap, **kwargs)

Replace the call to stackview.imshow() with modified_imshow() and adjust accordingly.

Regarding your suggestion for a scale bar argument, it's a great idea for an enhancement, but for immediate solutions, handling overlays externally before passing to imshow() might be the feasible path. I will tag @haesleinhuepf for any thoughts on integrating a scale bar feature.

Best,
git-bob

[haesleinhuepf]

git-bob implement this new if-statement in our imshow function in _imshow.py

[haesleinhuepf]

Hi @FloskelSalomon ,

I think the issue should be fixed in stackview 0.17.0. Please update your installation and let me know if the issue presists!

Thanks for the bug report!

Best,
Robert

[FloskelSalomon]

Hi @haesleinhuepf ,

Yes the error is fixed now with stackview 0.17.0.

Many thanks!

Best,
Florian