Scenery:Satellite and aerial image processing - Revision history

Thomas: Scenery:Landsat TM processing moved to Scenery:Satellite and aerial image processing

2008-12-19T20:57:18Z

Scenery:Landsat TM processing moved to Scenery:Satellite and aerial image processing

← Older revision	Revision as of 20:57, 19 December 2008
(No difference)

Thomas: /* Histogram stretching */

2008-12-19T20:55:32Z

Histogram stretching

← Older revision		Revision as of 20:55, 19 December 2008
Line 132:		Line 132:
	Notice how most of the details on the mountain tops are lost without masking.		Notice how most of the details on the mountain tops are lost without masking.

	[[Image:Masked-~~Comp~~-5.png]]		[[Image:Masked-comp-5.png]]

	Step 1 : Check the original for highlights. Skip the process if not required.		Step 1 : Check the original for highlights. Skip the process if not required.

Thomas: /* Histogram stretching */

2008-12-19T20:55:06Z

Histogram stretching

← Older revision		Revision as of 20:55, 19 December 2008
Line 132:		Line 132:
	Notice how most of the details on the mountain tops are lost without masking.		Notice how most of the details on the mountain tops are lost without masking.

			[[Image:Masked-Comp-5.png]]

	Step 1 : Check the original for highlights. Skip the process if not required.		Step 1 : Check the original for highlights. Skip the process if not required.

Thomas: /* Histogram stretching */

2008-12-19T20:53:48Z

Histogram stretching

← Older revision		Revision as of 20:53, 19 December 2008
Line 123:		Line 123:

	Below is an example of over saturation. Notice how the buildings in a city become almost pure white.		Below is an example of over saturation. Notice how the buildings in a city become almost pure white.

			[[Image:Stretched-2.png]]

	The snowy European Alp mountains are a problem in particular so I had to make a work around solution. I hacked vips so that I can mask out all the pixels with a value above a certain threshhold. I end up with an image that has holes (black pixels) cut in it. Then I stretch the masked out images and recombined them with the originals in imagelinker. This way I retain the snowy detail but can now get nice green valleys. It's not perfect but take a look at the results below and you'll see that it is certainly worth the effort even though the solution is a real kludge.		The snowy European Alp mountains are a problem in particular so I had to make a work around solution. I hacked vips so that I can mask out all the pixels with a value above a certain threshhold. I end up with an image that has holes (black pixels) cut in it. Then I stretch the masked out images and recombined them with the originals in imagelinker. This way I retain the snowy detail but can now get nice green valleys. It's not perfect but take a look at the results below and you'll see that it is certainly worth the effort even though the solution is a real kludge.

Thomas: /* Working with Landsat ETM+ imagery */

2008-12-19T20:52:56Z

Working with Landsat ETM+ imagery

← Older revision		Revision as of 20:52, 19 December 2008
Line 112:		Line 112:
	To do this load the image in OpenEV and run the Histogram Enhance tool. It may take a while to bring up a window so be patient - it's busy analysing the entire image in the background.		To do this load the image in OpenEV and run the Histogram Enhance tool. It may take a while to bring up a window so be patient - it's busy analysing the entire image in the background.
	Once the histogram tool has loaded select "linear" from the combo box and you should see an noticable improvement in the color and brightness like in the image below.		Once the histogram tool has loaded select "linear" from the combo box and you should see an noticable improvement in the color and brightness like in the image below.

			[[Image:Stretched-1.png]]

	Try playing around with the settings to get a feel of how the histogram stretching works. You can set stretch amount for each channel by clicking on the channels with the left and right mouse buttons. Just be careful to not over do the stretching or to make one band too different from the others otherwise you'll end with some very unatural looking images.		Try playing around with the settings to get a feel of how the histogram stretching works. You can set stretch amount for each channel by clicking on the channels with the left and right mouse buttons. Just be careful to not over do the stretching or to make one band too different from the others otherwise you'll end with some very unatural looking images.

Thomas: /* Histogram stretching */

2008-12-19T20:51:47Z

Histogram stretching

← Older revision		Revision as of 20:51, 19 December 2008
Line 106:		Line 106:
	At this stage you should have a rather dull looking image that is pan sharpened.		At this stage you should have a rather dull looking image that is pan sharpened.
	Below is an example of a pan-sharpened image in South Eastern France.		Below is an example of a pan-sharpened image in South Eastern France.

			[[image:Unstretched.png]]

	What we want to do increase the brightness and contrast of the image.		What we want to do increase the brightness and contrast of the image.

Thomas: /* Pan-sharpening the RGB composite */

2008-12-19T15:09:53Z

Pan-sharpening the RGB composite

← Older revision		Revision as of 15:09, 19 December 2008
Line 95:		Line 95:

	Below are two images that show what happens when you don't blur a noisy panchromatic band before a fusion process.		Below are two images that show what happens when you don't blur a noisy panchromatic band before a fusion process.

			[[Image:Sharp_comp_2.png]]

			[[Image:Sharp_comp_3.png]]

	As can be seen above using an unblurred, noisy panchromatic band creates a barely perceivable improvement in sharpness at the expense of a lot of noise when used in a fusion process.		As can be seen above using an unblurred, noisy panchromatic band creates a barely perceivable improvement in sharpness at the expense of a lot of noise when used in a fusion process.

Thomas: /* Pan-sharpening the RGB composite */

2008-12-19T15:09:03Z

Pan-sharpening the RGB composite

← Older revision		Revision as of 15:09, 19 December 2008
Line 87:		Line 87:

	Below are some comparisons after some histogram stretching (to make things look nicer).		Below are some comparisons after some histogram stretching (to make things look nicer).

			[[Image:Sharp_comp_1.png]]

			[[Image:Sharp_comp_4.png]]

	Notice how the edges of the dam and the roads are much more crisp. You can also see that the pan-sharpened image contains a bit more noise but this is not a problem because it will be smoothed out by OpenGL in the simulator.		Notice how the edges of the dam and the roads are much more crisp. You can also see that the pan-sharpened image contains a bit more noise but this is not a problem because it will be smoothed out by OpenGL in the simulator.

Thomas: /* Pan-sharpening the RGB composite */

2008-12-19T15:07:02Z

Pan-sharpening the RGB composite

← Older revision		Revision as of 15:07, 19 December 2008
Line 72:		Line 72:

	Below is an example of the original pan band versus a blurred pan band.		Below is an example of the original pan band versus a blurred pan band.

	[[Image:Pan-blurred.png]]		[[Image:Pan-blurred.png]]

Thomas: /* Pan-sharpening the RGB composite */

2008-12-19T15:06:44Z

Pan-sharpening the RGB composite

← Older revision		Revision as of 15:06, 19 December 2008
Line 67:		Line 67:

	To save and restore the GeoTIFF meta tags use listgeo and geotifcp (both part of the Open Source libgeotiff).		To save and restore the GeoTIFF meta tags use listgeo and geotifcp (both part of the Open Source libgeotiff).
	Example :		Example :
	listgeo original-pan-band.tif > pan_meta.txt		listgeo original-pan-band.tif > pan_meta.txt
	geotifcp -g pan_meta.txt modified-pan-band.tif output-pan-band-blurred.tif		geotifcp -g pan_meta.txt modified-pan-band.tif output-pan-band-blurred.tif

	Below is an example of the original pan band versus a blurred pan band.		Below is an example of the original pan band versus a blurred pan band.
			[[Image:Pan-blurred.png]]

	Notice how the noise is far less noticeable at the expense of a bit of sharpness.		Notice how the noise is far less noticeable at the expense of a bit of sharpness.
Line 91:		Line 92:

	As can be seen above using an unblurred, noisy panchromatic band creates a barely perceivable improvement in sharpness at the expense of a lot of noise when used in a fusion process.		As can be seen above using an unblurred, noisy panchromatic band creates a barely perceivable improvement in sharpness at the expense of a lot of noise when used in a fusion process.

	==Histogram stretching==		==Histogram stretching==