Visualizing noisy images

Anthony — Wed, 13 May 2009 11:51:43 +0000

You have an image. Each pixel has a value with some uncertainty. How do you visualize the uncertainty in each pixel? Like this:

Here's the Python code

import numpy as np
from matplotlib import pyplot as plt
 
class FlickerImage(object):
    def __init__(self, im, err):
        self.im = im.copy()
        self.err = err.copy()
        finite = np.isfinite(self.im + self.err)
        self.vmin = (self.im - 2 * self.err)[finite].min()
        self.vmax = (self.im + 2 * self.err)[finite].max()
        self.im[np.invert(finite)] = self.vmax
        self.err[np.invert(finite)] = 0
    def flicker(self):
        fg = plt.imshow(np.zeros(self.im.shape),
                        interpolation='nearest',
                        vmin=self.vmin,
                        vmax=self.vmax)
        while True:
            ran = np.random.normal(size=im.shape)
            fg.set_data(im + err * ran)
            plt.draw()

And here's an example script:

import pyfits
f = pyfits.open('file.fits')
im = f["IMAGE"].data
err = f["ERROR"].data
flicker_image = FlickerImage(im, err)
flicker_image.flicker()

Python, FITS and DS9

Anthony — Wed, 01 Apr 2009 13:06:34 +0000

Here's an easy way to display FITS images (or any array) in DS9 using Python (with PyFITS, NumPy and Numdisplay, which is part of stsci_python). First launch DS9, then in Python:

import numdisplay
import pyfits
arr = pyfits.getdata('file.fits')
numdisplay.display(arr)

Easy!

Alternatively, the Kapteyn package seems excellent, and uses Python's matplotlib for displaying images. ~~It requires WCSLIB to run, though, so the installation process is a bit longer~~ [update: apparently that bit is no longer true - 17 Oct 2011].

A third option is to use python-sao:

import pysao
import pyfits
ds9 = pysao.ds9()
f = pyfits.open('file.fits')
ds9.view(f[0])

Easy again! And the WCS information is preserved, which doesn't seem to be the case with Numdisplay.

Anthony Smith's Research Blog » visualization

Visualizing noisy images

Python, FITS and DS9