In a matplotlib figure window (with imshow), how c

问题:

This question already has an answer here:

• Interactive pixel information of an image in Python? 4 answers

I am using ipython with matplotlib, and I show images in this way:

(started up with: ipython --pylab)

figure()  
im = zeros([256,256]) #just a stand-in for my real images   
imshow(im)

Now, as I move the cursor over the image, I see the location of the mouse displayed in the lower left corner of the figure window. The numbers displayed are x = column number, y = row number. This is very plot-oriented rather than image-oriented. Can I modify the numbers displayed?

1. My first choice would be to display x = row number*scalar, y = column number*scalar
2. My second choice would be to display x = row number, y = column number
3. My third choice is to not display the numbers for the mouse location at all

Can I do any of these things? I'm not even sure what to call that little mouse-over test display widget. Thanks!

回答1:

You can do this quite simply on a per axis basis by simply re-assigning format_coord of the Axes object, as shown in the examples.

format_coord is any function which takes 2 arguments (x,y) and returns a string (which is then displayed on the figure.

If you want to have no display simply do:

ax.format_coord = lambda x, y: ''

If you want just the row and column (with out checking)

scale_val = 1
ax.format_coord = lambda x, y: 'r=%d,c=%d' % (scale_val * int(x + .5), 
                                             scale_val * int(y + .5))

If you want to do this on every iimage you make, simply define the wrapper function

def imshow(img, scale_val=1, ax=None, *args, **kwargs):
    if ax is None:
         ax = plt.gca()
    im = ax.imshow(img, *args, **kwargs)
    ax.format_coord = lambda x, y: 'r=%d,c=%d' % (scale_val * int(x + .5), 
                                             scale_val * int(y + .5))
    ax.figure.canvas.draw()
    return im

which with out much testing I think should more-or-less be drop-in replacement for plt.imshow

回答2:

Yes, you can. But it's harder than you'd think.

The mouse-tracking label you see is generated by calls to matplotlib.axes.Axes.format_coord in response to mouse tracking. You have to create your own Axes class (overriding format_coord to do what you want it to do), then instruct matplotlib to use it in place of the default one.

Specifically:

Make your own Axes subclass

from matplotlib.axes import Axes
class MyRectilinearAxes(Axes):
    name = 'MyRectilinearAxes'
    def format_coord(self, x, y):
        # Massage your data here -- good place for scalar multiplication
        if x is None:
            xs = '???'
        else:
            xs = self.format_xdata(x * .5)
        if y is None:
            ys = '???'
        else:
            ys = self.format_ydata(y * .5)
        # Format your label here -- I transposed x and y labels
        return 'x=%s y=%s' % (ys, xs)

Register your Axes subclass

from matplotlib.projections import projection_registry
projection_registry.register(MyRectilinearAxes)

Create a figure and with your custom axes

figure()
subplot(111, projection="MyRectilinearAxes")

Draw your data as before

im = zeros([256,256]) #just a stand-in for my real images
imshow(im)