Image processing with Python and SciPy - Revision history

WikiSysop at 17:54, 14 April 2018

2018-04-14T17:54:48Z

← Older revision		Revision as of 17:54, 14 April 2018
Line 57:		Line 57:
	pip install scikit-image		pip install scikit-image

	should work. The package usually requires local compilation of the code when installed this way. Qt bindings are provided by PySide or by Qt, depending on licensing requirements. PySide is less restrictive.		should work. The package usually requires local compilation of the code when installed this way. Qt bindings are provided by PyQt5, PySide or by qtpy, depending on licensing and software requirements. PySide is less restrictive than PyQt and qtpy is an abstraction that draws on both.

	pip install PySide		pip install PySide

WikiSysop at 13:48, 14 April 2018

2018-04-14T13:48:29Z

← Older revision		Revision as of 13:48, 14 April 2018
Line 272:		Line 272:


	=== Reading and ~~Displaying~~ a FITS File in Python ===		=== Reading and Writing a FITS File in Python ===

	There are many image display tools for astronomy, and perhaps the most widely used is [http://hea-www.harvard.edu/RD/ds9/site/Home.html ds9] which is available for Linux, MacOS, and Windows, as well as in source code. It is always useful to have a version on your computer when you are working with FITS images. A more versatile Java platform for astronomical image viewing that also does and processing' is now widely used for precision astronomical photometry where interactive analysis is needed. AstroImageJ is free and simple to install on most computers, and because it is also a powerful processor, for many purposes it is an all-in-one tool. If you are working with FITS images, this is highly recommended:		There are many image display tools for astronomy, and perhaps the most widely used is [http://hea-www.harvard.edu/RD/ds9/site/Home.html ds9] which is available for Linux, MacOS, and Windows, as well as in source code. It is always useful to have a version on your computer when you are working with FITS images. A more versatile Java platform for astronomical image viewing that also does and processing' is now widely used for precision astronomical photometry where interactive analysis is needed. AstroImageJ is free and simple to install on most computers, and because it is also a powerful processor, for many purposes it is an all-in-one tool. If you are working with FITS images, this is highly recommended:

WikiSysop at 13:43, 14 April 2018

2018-04-14T13:43:24Z

← Older revision		Revision as of 13:43, 14 April 2018
Line 4:		Line 4:
	== Pillow - An Imaging Library ==		== Pillow - An Imaging Library ==

	The Python Imaging Library was developed for Python 2.x and provided functions to manipulate images, including reading, modifying and saving in various standard image formats in a package called "PIL". With the coming of age of Python 3.x, a fork of the older version has evolved that is more suited for the new technologies and is in a package called "Pillow". It continues to improve, and the features described here are tested with "Pillow 5.1" and Python 3.6 as of April 2018. Pillow will probably be on any packaged distribution of Python 3, or it may be installed with (note the capital "P")		The Python Imaging Library (PIL) was developed for Python 2.x and provided functions to manipulate images, including reading, modifying and saving in various standard image formats in a package called "PIL". With the coming of age of Python 3.x, a fork of the older version has evolved that is more suited for the new technologies and is in a package called "Pillow". It continues to improve, and the features described here are tested with "Pillow 5.1" and Python 3.6 as of April 2018. Pillow will probably be on any packaged distribution of Python 3, or it may be installed with (note the capital "P")

	pip install Pillow		pip install Pillow

WikiSysop at 13:42, 14 April 2018

2018-04-14T13:42:25Z

← Older revision		Revision as of 13:42, 14 April 2018
Line 8:		Line 8:
	pip install Pillow		pip install Pillow

	Pillow includes the basics of image processing. with functions that are documented by the developers in a [https://pillow.readthedocs.io/en/5.1.x/handbook/index.html handbook] describing the methods and giving some examples. Pillow uses the same "namespace" as PIL and older code should work, perhaps with a few modifications to allow for recent developments. Most import for us, Pillow has routines to read and write conventional image formats. Once an image has been read into a numpy array, the full power of Python is available to process it, and we can turn to Pillow again to save a processed image in as png or jpg or another format. FITS ~~format~~ files used for astronomy should be managed with astropy or pyfits.		Pillow includes the basics of image processing. with functions that are documented by the developers in a [https://pillow.readthedocs.io/en/5.1.x/handbook/index.html handbook] describing the methods and giving some examples. Pillow uses the same "namespace" as PIL and older code should work, perhaps with a few modifications to allow for recent developments. Most import for us, Pillow has routines to read and write conventional image formats. Once an image has been read into a numpy array, the full power of Python is available to process it, and we can turn to Pillow again to save a processed image in png or jpg or another format. Flexibile Image Transport System (FITS) files used for astronomy should be managed with astropy or pyfits.

	As a simple starting example, suppose you have an image that was taken with the camera turned so that "up" is to the side when the image is displayed. Here's how you would rotate ~~an image~~ 90 degrees.		As a simple starting example, suppose you have an image that was taken with the camera turned so that "up" is to the side when the image is displayed. Here's how you would read it, rotate it 90 degrees, and write it out again using Pillow.

	import os		import os

WikiSysop at 13:39, 14 April 2018

2018-04-14T13:39:06Z

← Older revision		Revision as of 13:39, 14 April 2018
Line 1:		Line 1:
	Given that NumPy provides multidimensional arrays, and that there is core support through the Python Imaging Library and Matplotlib to display images and manipulate images in the Python environment, it's easy to take the next step and combine these for scientific image processing. As part of our [http://prancer.physics.louisville.edu/astrowiki/index.php/Python_for_Physics_and_Astronomy short course on Python for Physics and Astronomy] we begin by exploring how Python handles image input and output through pillow, scikit-image, and pyfits. Once loaded, an image may be processed using library routines or by mathematical operations that would take advantage of the speed and conciseness of numpy and scipy.		Given that NumPy provides multidimensional arrays, and that there is core support through the Python Imaging Library and Matplotlib to display images and manipulate images in the Python environment, it's easy to take the next step and combine these for scientific image processing. As part of our [http://prancer.physics.louisville.edu/astrowiki/index.php/Python_for_Physics_and_Astronomy short course on Python for Physics and Astronomy] we begin by exploring how Python handles image input and output through pillow, scikit-image, and pyfits. Once loaded, an image may be processed using library routines or by mathematical operations that would take advantage of the speed and conciseness of numpy and scipy. Some of the resources mentioned here require Python >3.4, and at this time Python 3.6 is the current one.


	== ~~Python~~ Imaging Library ~~- Pillow~~ ==		== Pillow - An Imaging Library ==

	The Python Imaging Library was developed for Python 2.x and provided functions to manipulate images, including reading, modifying and saving in various standard image formats in a package called "PIL". With the coming of age of Python 3.x, a fork of the older version has evolved that is more suited for the new technologies and is in a package called "Pillow". It continues to improve, and the features described here are tested with "Pillow 5.1" and Python 3.6 as of April 2018. Pillow will probably be on any packaged distribution of Python 3, or it may be installed with (note the capital "P")		The Python Imaging Library was developed for Python 2.x and provided functions to manipulate images, including reading, modifying and saving in various standard image formats in a package called "PIL". With the coming of age of Python 3.x, a fork of the older version has evolved that is more suited for the new technologies and is in a package called "Pillow". It continues to improve, and the features described here are tested with "Pillow 5.1" and Python 3.6 as of April 2018. Pillow will probably be on any packaged distribution of Python 3, or it may be installed with (note the capital "P")
Line 8:		Line 8:
	pip install Pillow		pip install Pillow

	Pillow includes the basics of image processing. with functions that are documented by the developers in a [https://pillow.readthedocs.io/en/5.1.x/handbook/index.html handbook] describing the methods and giving some examples. Pillow uses the same "namespace" as PIL and older code should work, perhaps with a few modifications to allow for recent developments.		Pillow includes the basics of image processing. with functions that are documented by the developers in a [https://pillow.readthedocs.io/en/5.1.x/handbook/index.html handbook] describing the methods and giving some examples. Pillow uses the same "namespace" as PIL and older code should work, perhaps with a few modifications to allow for recent developments. Most import for us, Pillow has routines to read and write conventional image formats. Once an image has been read into a numpy array, the full power of Python is available to process it, and we can turn to Pillow again to save a processed image in as png or jpg or another format. FITS format files used for astronomy should be managed with astropy or pyfits.

	As a simple starting example, suppose you have an image that was taken with the camera turned so that "up" is to the side when the image is displayed. Here's how you would rotate an image 90 degrees.		As a simple starting example, suppose you have an image that was taken with the camera turned so that "up" is to the side when the image is displayed. Here's how you would rotate an image 90 degrees.
Line 45:		Line 45:

	Processing is not limited to "PNG" files, though that file type is preferred because it is not a lossy storage option. Python reads and writes "jpg" files too. While PIL provides some essential functionality, for more practical uses in astronomy we need to read Flexible Image Transport or "FITS" files, and to enable numerical work on images in SciPy. Many of the processing functions you will find in Python Imaging Library (PIL) are also available in SciPy where we have precise mathematical control over their definitions and operation. Some more advanced techiques are available in SciPy too, courtesy of researchers who have contributed to SciKit as we will see.		Processing is not limited to "PNG" files, though that file type is preferred because it is not a lossy storage option. Python reads and writes "jpg" files too. While PIL provides some essential functionality, for more practical uses in astronomy we need to read Flexible Image Transport or "FITS" files, and to enable numerical work on images in SciPy. Many of the processing functions you will find in Python Imaging Library (PIL) are also available in SciPy where we have precise mathematical control over their definitions and operation. Some more advanced techiques are available in SciPy too, courtesy of researchers who have contributed to SciKit as we will see.

			Python's core routines dependent on matplotlib may be used to display an image, but these are designed for graphics, and limited by the constraints of the matplotlib interface. With a little effort there are better choices.

WikiSysop: /* SciKit Image */

2018-04-14T05:53:17Z

SciKit Image

← Older revision		Revision as of 05:53, 14 April 2018
Line 55:		Line 55:
	pip install scikit-image		pip install scikit-image

	should work. The package usually requires local compilation of the code when installed this way.		should work. The package usually requires local compilation of the code when installed this way. Qt bindings are provided by PySide or by Qt, depending on licensing requirements. PySide is less restrictive.

			pip install PySide


	While scipy has included an image reader and writer, as of April 2018 this function is deprecated in the base code and rather than use pillow, we can turn to scikit-image. The module to read and write image is skimage.io		While scipy has included an image reader and writer, as of April 2018 this function is deprecated in the base code and rather than use pillow, we can turn to scikit-image. The module to read and write image is skimage.io
Line 71:		Line 74:

	and imlibs['pil'] will list the functions that the Python imaging library provides. The package tries the libraries in order until it finds one that works:		and imlibs['pil'] will list the functions that the Python imaging library provides. The package tries the libraries in order until it finds one that works:

	myimage = skimage.io.imread(filename).astype( np.float32)		myimage = skimage.io.imread(filename).astype( np.float32)

	will read an image and return a numpy array which by default will be an RGB image if the file is a png file, for example. A greyscale image image be specified by including as_grey=True as an argument. A numpy image has a shape that for color has 3 values in each pixel		will read an image and return a numpy array which by default will be an RGB image if the file is a png file, for example. A greyscale image image be specified by including as_grey=True as an argument. A numpy image has a shape that for color has 3 values in each pixel

	print(myimage.shape)		print(myimage.shape)
	(498, 680, 3)		(498, 680, 3)

	as an example for an RGB image, or		as an example for an RGB image, or

	(498,680)		(498,680)

	for a gray scale image. Since numpy by default would store into a 64-bit float and matplotlib (the default display for skimage) requires 32-bit, we specify loading into a 32 bit array while planning ahead to seeing the result.		for a gray scale image. Since numpy by default would store into a 64-bit float and matplotlib (the default display for skimage) requires 32-bit, we specify loading into a 32 bit array while planning ahead to seeing the result.


	Images may be saved:		Images may be saved:

	skimage.io.imsave(filename, nparray)		skimage.io.imsave(filename, nparray)

	and the file type is determined by file name extension.		and the file type is determined by file name extension.

	Images may be displayed, but it takes two steps		Images may be displayed, but it takes two steps
	skimage.io.imshow(~~nparray~~)
			skimage.io.imshow(myimage)
	skimage.io.show()		skimage.io.show()
	~~when invoking the default matplotlib plugin. The display will look like one created by pyplot.~~

			when invoking the default matplotlib plugin. The display will look like one created by pyplot. There is a simpler, viewer module too, without pyplot toolbar.

			import skimage.viewer

			viewer = skimage.viewer.viewers.ImageViewer(myimage)
			viewer.show


Line 135:		Line 151:

	[[File:test_m34_stars.png \| center \| 600px]]		[[File:test_m34_stars.png \| center \| 600px]]


	== Images with NumPy and SciPy ==		== Images with NumPy and SciPy ==

WikiSysop at 05:33, 14 April 2018

2018-04-14T05:33:01Z

← Older revision		Revision as of 05:33, 14 April 2018
Line 60:		Line 60:

	import skimage.io		import skimage.io
			import numpy as np

	and the command		and the command
Line 70:		Line 71:

	and imlibs['pil'] will list the functions that the Python imaging library provides. The package tries the libraries in order until it finds one that works:		and imlibs['pil'] will list the functions that the Python imaging library provides. The package tries the libraries in order until it finds one that works:
			myimage = skimage.io.imread(filename).astype( np.float32)
	myimage = skimage.io.imread(filename)

	will read an image and return a numpy array which by default will be an RGB image if the file is a png file, for example. A greyscale image image be specified by including as_grey=True as an argument. A numpy image has a shape that for color has 3 values in each pixel		will read an image and return a numpy array which by default will be an RGB image if the file is a png file, for example. A greyscale image image be specified by including as_grey=True as an argument. A numpy image has a shape that for color has 3 values in each pixel

	print(myimage.shape)		print(myimage.shape)
	(498, 680, 3)		(498, 680, 3)
			as an example for an RGB image, or
			(498,680)
			for a gray scale image. Since numpy by default would store into a 64-bit float and matplotlib (the default display for skimage) requires 32-bit, we specify loading into a 32 bit array while planning ahead to seeing the result.


	~~as an example for an RGB image~~, or		Images may be saved:
			skimage.io.imsave(filename, nparray)
			and the file type is determined by file name extension.

	(~~498,680~~)		Images may be displayed, but it takes two steps
			skimage.io.imshow(nparray)
			skimage.io.show()
			when invoking the default matplotlib plugin. The display will look like one created by pyplot.

	~~for a gray scale image. Images may be saved and displayed too:~~

	~~skimage.io.imsave(filename,nparray)~~
	~~skimage.io.imshow(nparray, plugin=selected_plugin)~~

WikiSysop at 05:00, 14 April 2018

2018-04-14T05:00:36Z

Show changes

WikiSysop at 05:51, 10 April 2018

2018-04-10T05:51:57Z

← Older revision		Revision as of 05:51, 10 April 2018
Line 582:		Line 582:


	== ~~Examples~~ ==		== AstroImageJ and Alsvid ==

			We have developed a collection of Python routines to do many of the routine astronomical image processing tasks such as dark subtraction, flat fielding, co-addition, and FITS header management through PyFITS and PyWCS. The current version of the "Alsvid" package is available for download:


	~~For examples of Python illustrating image processing, see the~~ [http://~~prancer~~.~~physics~~.louisville.edu/~~astrowiki~~/index.~~php/Python_examples examples~~] ~~section.~~		[http://www.astro.louisville.edu/software/alsvid/index.html Alsvid]


	~~== AstroImageJ and Alsvid ==~~

	We have developed a collection of Python routines to do many of the routine astronomical image processing tasks such as dark subtraction, flat fielding, co-addition, and FITS header management through PyFITS and PyWCS. The current version of the "Alsvid" package is available for download:		== Examples ==

			For examples of Python illustrating image processing, see the [http://prancer.physics.louisville.edu/astrowiki/index.php/Python_examples examples] section.

	~~[http://www.astro.louisville.edu/software/alsvid/index.html Alsvid]~~

WikiSysop at 05:44, 10 April 2018

2018-04-10T05:44:51Z

← Older revision		Revision as of 05:44, 10 April 2018
Line 156:		Line 156:


	== Astronomical FITS Files ~~with PyFITS~~ ==		== Astronomical FITS Files ==

	~~[http://www~~.~~stsci.edu/institute/software_hardware/pyfits PyFITS] is available from the Space Telescope Science Institute~~, and ~~can be added easily~~ to a ~~Python installation that already has NumPy and SciPy~~. As of ~~January 2013~~, the ~~current version 3~~.~~1.1 of PyFITS supports all~~ the ~~functions needed~~ to ~~manage~~ image and ~~table data in~~ the ~~standard Flexibile~~ Image Transport System (FITS) files of astronomy.		Astronomical image data are potentially complex and rich, for which quantitative structures have been developed to standardize lossless storage of the data along with the metadata that describe its origin and previous processing. While photographic images are often only 8 bits deep, mixed with red, green and blue in a single image, and compressed to reduce file size, astronomical images are 16 or 32 bits deep in a single color. Until recently, the file sizes needed for astronomy were unrivaled by commodity cameras, but in todays market of megapixel imaging on cell phones, the camera in your pocket also produces very large rich images. The difference in handling them is that for science we want to preserve the data without loss, quantitatively calibrate and measure the flux from the source and map that back to a specific angle in space, while for art or even for some academic uses, it is the beauty, color, and highlights shown in the image display that are important. Commodity images are saved usually in compressed formats such as JPG, or uncompressed TIFF or proprietary binary formats. For astronomy and other quantitative imaging work, the [https://fits.gsfc.nasa.gov/fits_primer.html Flexible Image Transport System (or FITS)] format is almost universal. It includes the image data, and a header describing the data. FITS files may also be tables of data, or a cube of images in sequence. The standards developed for creating these files are slowing evolving as the needs of big data in astronomy have grown. Programmers and scientists at NASA, the Space Telescope Science Institute, and the academic community at large are contributing to libraries that enable reading, processing, and saving FITS files in Python, as well as in C, and Fortran.

	Once a FITS file has been read, the header is accessible ~~for reading~~, and the image data are ~~available~~ in a NumPy array. With ~~PyFITS~~ and ~~NumPY~~ you can read, extract information, modify, display and save image data.		Once a FITS file has been read, the header its accessible as a Python dictionary of the data contents, and the image data are in a NumPy array. With Python using NumPy and SciPy you can read, extract information, modify, display, create and save image data.


	=== Displaying a FITS File in Python ===		=== Reading and Displaying a FITS File in Python ===

	There are many image display tools for astronomy, and perhaps the most widely used is [http://hea-www.harvard.edu/RD/ds9/site/Home.html ds9] which is available for Linux, MacOS, and Windows, as well as in source code. It is useful to have a version on your computer when you are working with FITS images. A more versatile Java platform for astronomical image viewing ''and processing'' is now widely used for precision astronomical photometry where interactive analysis is needed. AstroImageJ is free and simple to install on most computers.		There are many image display tools for astronomy, and perhaps the most widely used is [http://hea-www.harvard.edu/RD/ds9/site/Home.html ds9] which is available for Linux, MacOS, and Windows, as well as in source code. It is always useful to have a version on your computer when you are working with FITS images. A more versatile Java platform for astronomical image viewing that also does and processing' is now widely used for precision astronomical photometry where interactive analysis is needed. AstroImageJ is free and simple to install on most computers, and because it is also a powerful processor, for many purposes it is an all-in-one tool. If you are working with FITS images, this is highly recommended:

	[http://www.astro.louisville.edu/software/astroimagej/index.html AstroImageJ] website and program source		[http://www.astro.louisville.edu/software/astroimagej/index.html AstroImageJ] website and program source
Line 172:		Line 172:


	However, sometimes we want to view ~~an image~~ while ~~we are~~ processing it in Python~~, and for that we could use a~~ routine ~~such as this one. It shows by example~~ how to ~~load an image that is stored in~~ a ~~[ https://fits.gsfc.nasa.gov/fits_primer.html "Flexible Image Transport System " or~~ FITS] file ~~format~~, the ~~standard for astronomical and many scientific imaging applications~~. ~~This format~~ is ~~non-trivial to read and write, and libraries for C, Fortran, and Python are available and frequently updated. The Python library appeared first as~~ "PyFITS" ~~and was~~ developed ~~by the~~ Space Telescope Science Institute. ~~It has been absorbed~~ into a larger package ~~for astronomical image processing, and now rather than install pyfits alone, we usually install~~ the ~~entire package called~~ astropy. It is part of the Enthought and Canopy distributions of Python. If you are using a system version of python, you may need to install it with pip		However, sometimes we want perform specialized work on image data, and to view it while processing it in Python. This routine demonstrates how to read a FITS file, inspect its header, and show the image on the computer display. It is dependent on the "PyFITS" library developed at Space Telescope Science Institute and incorporated into a larger package by the [http://www.astropy.org/ AstroPy Project]. AstroPy's library is part of the Enthought and Canopy distributions of Python. If you are using a system version of python, you may need to install it with pip

	pip install astropy		pip install astropy
Line 189:		Line 189:
	import matplotlib.pyplot		import matplotlib.pyplot

	~~Sometimes~~ submodules are not loaded with the larger package, ~~and~~ we ~~explictly~~ ask for pyplot.		Importing the FITS modules in this way makes the code backward compatible with the earlier versions of PyFITS. Also, since sometimes submodules are not loaded with the larger package, we explicitly ask for the io.fits components, and for pyplot.

	# Define a function for making a linear gray scale		# Define a function for making a linear gray scale