My previous article on this subject discussed some image processing techniques designed to produce visually pleasing color RGB composite images from Landsat 8 band file data. The techniques presented in that article proposed some basic image processing techniques like histogram stretching and gamma correction to transform Landsat 8 band composite RGB images to resemble the images from corresponding Landsat 7 data more closely.
After publishing the article, Mr. Guy Serbin suggested using reflectance data rather than digital numbers (DNs) in order to achieve the same thing. Mr. Binesh Maharjan kindly pointed me to the USGS web page that listed the transformation equations for computing Top of Atmosphere (TOA) reflectance and atmospheric radiance from Landsat 8 DNs. (The website can be found at the page entitled: Using the USGS Landsat 8 Product ).
I added these transformations to the PANCROMA™ preprocessing utility suite and conducted some additional experiments to see how the technique worked. In general, it worked very well indeed, producing results that were fairly close to the Landsat 7 target images.
PANCROMA TOA transformations must be made one Landsat 8 band file at a time. Open a band file by selecting 'File' | 'Open'. Then select 'PreProcess' | 'Compute Radiometric Corrections' | 'Compute Landsat 8 TOA Reflectance'. The TOA Reflectance Data Form will become visible. This form will ask you to input the Solar Elevation Angle from the metadata file that accompanies each Landsat 8 scene.
Note that unlike Landsat 7, the conversion coefficients are the same for each Landsat 8 band so it is not necessary to enter the band number. In addition, the conversion does not use the sun-earth distance (although this parameter is listed in the metadata). The reference Landsat 7 RGB image is shown below. The corner coordinates are:
ULX is: 339090
ULY is: -2540850
LRX is: 365850
LRY is: -2564040
Landsat 7 bands 1, 2, 3 RGB color composite.
The next image has been composited from the unprocessed Landsat 8 bands 2, 3 and 4. This image does not resemble the Landsat 7 version very closely.
Landsat 8 bands 2, 3, 4 RGB color composite, unprocessed.
The next image has been composited from scaled TOA reflectances prepared using the new PANCROMA™ utility. This image is much closer to the Landsat 7 reference image.
Landsat 8 bands 2, 3, 4 RGB color composite, TOA reflectance bands.
Next, I performed a haze reduction using the Haze Optimized Transform method.
Landsat 8 bands 2, 3, 4 RGB color composite, TOA reflectance bands, HOT haze reduction.
I then performed an histogram stretch on the image. This has improved the image but introduced some blue tones.
Landsat 8 bands 2, 3, 4 RGB color composite, TOA reflectance bands, HOT haze reduction, histogram stretch.
Finally, I performed a gamma correction, brightening the image and balancing the blue tones a bit. (It looks like I over corrected a bit, but you can get the idea of what is possible from the image).
Landsat 8 bands 2, 3, 4 RGB color composite, TOA reflectance bands, HOT haze reduction, histogram stretch, gamma corrected.
I am not sure which is the "best" RGB color composite image. The image developed from the unprocessed reflectance data is very similar to the Landsat 7 reference image. I like the fully processed (haze reduced, stretched and gamma corrected) as it shows some detail not quite so visible in the other images. However it is very subjective. One main point is that Guy's suggestion worked very well, suggesting that TOA reflectance conversion will be a much more widely used technique for Landsat 8 data than for Landsat 7. Another point is that Landsat 8 images can be made a lot more visually pleasing (and can reveal more information) with a bit of image processing. You can investigate more Landsat satellite image processing techniques at www.PANCROMA.com menu selection 'White Papers'