How do I remove an image overlay in Matplotlib?

Using matplotlib and python, I have a grey-scale image of labeled objects, on which I want to draw a homogeneously coloured overlay image with a position and shape based on a changeable input parameter - an object identifier. Basically an outline and enhancement of on of the objects in the image.

I can generate the overlay, and re-generate it correctly (I think) every time the input value changes. But I don't know how to clear the previous overlay before drawing the new one. So, in the end, the grey-scale image is overlaid with multiple overlays.

This is what I have tried, and it doesn't work. 'overlay',and 'object_data' are defined and used in the calling function:

    def overlay_object(object_num):
        try:
            overlay
        except NameError:
            # nothing
            pass
        else:
            # remove old overlay
            for handle in overlay:
                handle.remove()

        # color the selected object
        componentMask = (object_data == object_num) 
        masked = ma.masked_where(componentMask == 0, componentMask)
        overlay = ax.imshow(masked, 'jet', interpolation='none', alpha=0.5)
        return overlay

Edit:

This is the creation of the grey-scale image in the main program:

fig, ax = plt.subplots()
ax.imshow(object_data, cmap='gray') 
ax.axis('off')

If you’re trying to update overlay on a grayscale without accumulating overlays, you should use this approach:

import matplotlib.pyplot as plt
import numpy as np
import numpy.ma as ma

def create_interactive_overlay(object_data):
    """
    Creates a figure with a grayscale base image and functions to update overlays.
    
    Parameters:
    object_data : numpy.ndarray
        The labeled image data where each object has a unique integer value
    
    Returns:
    fig : matplotlib.figure.Figure
        The figure object
    update_overlay : function
        Function to call to update the overlay
    """
    # Create the figure and base image
    fig, ax = plt.subplots()
    ax.imshow(object_data, cmap='gray')
    ax.axis('off')
    
    # Initialize overlay as None
    overlay_artist = [None]  # Use list to allow modification in nested function
    
    def update_overlay(object_num):
        """
        Updates the overlay to highlight a specific object number.
        
        Parameters:
        object_num : int
            The object identifier to highlight
        """
        # Remove existing overlay if it exists
        if overlay_artist[0] is not None:
            overlay_artist[0].remove()
        
        # Create mask for selected object
        component_mask = (object_data == object_num)
        masked = ma.masked_where(component_mask == 0, component_mask)
        
        # Create new overlay
        overlay_artist[0] = ax.imshow(masked, cmap='jet', 
                                    interpolation='none', alpha=0.5)
        
        # Redraw the figure
        fig.canvas.draw_idle()
    
    return fig, update_overlay

# Example usage:
"""
# Create sample data
object_data = np.zeros((100, 100))
object_data[20:40, 20:40] = 1
object_data[60:80, 60:80] = 2

# Create interactive figure
fig, update_overlay = create_interactive_overlay(object_data)

# Update overlay for different objects
update_overlay(1)  # Highlight object 1
plt.pause(1)  # Pause to see the change
update_overlay(2)  # Highlight object 2

plt.show()
"""

In the above solution, the overlay management is handled by:

• Keeping track of the overlay artist using a list (to allow modification in the nested function)
• Properly removing the previous overlay before creating a new one
• Using draw_idle() to update the figure

The code is structured to separate the setup from the update functionality:

• create_interactive_overlay handles the initial setup
• update_overlay handles the dynamic updates

To use this in your code, you would do something like this:

fig, update_overlay = create_interactive_overlay(object_data)

# When you want to highlight a different object:
update_overlay(new_object_number)

Hope this helps