Image Fusion: Comparing PANSHARP and MRAFUSION

Image fusion is the process by which multiple images of the same scene are combined into a single image containing all the necessary information from the bigger dataset. It helps consolidate the data into a more understandable and informative product.

It is important to note that the Rotation Values of the PAN (Panchromatic) and MS (Multi-Spectral) images affect the performance and results of the PANSHARP and MSFUSION algorithms. While MRAFUSION does not accept MS and PAN images with different rotation values, PANSHARP can tolerate these images; however, the result will not be of good quality. The rotation value of your images can be found in the file properties in Focus:

1. In Focus, navigate to the Files tab
2. Right-click the file > Properties... > Projection

Additionally, the PAN and MS imagery should be coregistered before running the fusion algorithms to avoid blurry image results. The images can be coregistered using the Super Registration Algorithm. To learn more about Super Registration, please review the Super Registration in CATALYST Professional tutorial.

CATALYST Professional offers various image fusion techniques such as FUSE, PANSHARP, MRAFUSION, and so on.

PANSHARP

The PANSHARP algorithm increases the resolution of multispectral images by using image fusion algorithms with a high-resolution panchromatic image, and produces the best colour representation of the original images. The algorithms used address colour distortion and operator/data dependency issues from existing fusion algorithms. To learn more about the PANSHARP algorithm, visit the CATALYST Professional Help Documentation.

Pansharpening of images can be done through our various tools - Focus (via the Algorithm Librarian), OrthoEngine, Python, and EASI.

Note: Although PANSHARP accepts images with different rotation values, the result will not be of good quality.

The code block below is an example of the PANSHARP algorithm through Python.

from pci.pansharp import pansharp

fili = "toronto_ms_demo.pix"
dbic = [1,2,3,4]
dbic_ref = dbic	        # Reference images are also the image to be sharpened
fili_pan = "toronto_pan_demo.pix"
dbic_pan = [1]
srcbgd	= "ANY,0"	# Zero-valued pixels in any input image
enhance	= "YES"	        # Apply the color enhancement operation
resample = "CUBIC"      # Uses cubic resampling when resampling MS image to high resolution
filo = "pansharp.pix"
dboc = [1,2,3,4]
ftype = "TIF"		# The output file will be in GeoTIFF format
foptions = "TILED512"	# Output file will be tiled into 512x512 pixel tiles
poption	= "AVER"	# Unweighted averaging resampling used to build pyramid overview images

pansharp (fili, dbic, dbic_ref, fili_pan, dbic_pan, srcbgd, enhance, resample, filo, dboc, ftype, foptions, poption)

MRAFUSION

Similar to the PANSHARP algorithm, MRAFUSION (Multi-Resolution Analysis Fusion) can also be used to increase the resolution of multispectral imagery by using a high-resolution panchromatic image. It uses wavelet decomposition, which allows the original image to be perfectly recovered from decomposed images regardless of the decomposition levels. To learn more about the MRAFUSION algorithm, visit the CATALYST Professional Help Documentation.

MRAFUSION can be performed through our various tools - Focus (via the Algorithm Librarian), OrthoEngine, Python, and EASI.

The code block below is an example of the MRAFUSION algorithm through Python.

from pci.mrafusion import mrafusion

fili = "toronto_ms_demo.pix"
dbic = [1,2,3,4]
fili_pan = "toronto_pan_demo.pix"
dbic_pan = [1]
srcbgd = "ANY,0"	# Exclude zero-valued pixels in any input image
levels = []	        # Use system-computed number of levels
radapt = "YES"	        # Apply radiometric adaptation
edgeshar = [1]          # No edge sharpening
filo = "pansharp.pix"
dboc = [1,2,3,4]
# to build pyramid-overview images
ftype = "TIF"		# Create output file in GeoTIFF format
foptions = "TILED512"	# Tile output file into tiles 512 x 512 pixels
poption = "AVER"	# Use unweighted averaging resampling

mrafusion (fili, dbic, fili_pan, dbic_pan, srcbgd, levels, radapt, edgeshar, filo, dboc, ftype, foptions, poption)

Comparison

Similarities

• Both create high-resolution colour images
• Both fuse black & white panchromatic and multispectral colour images
• Both can fuse images from the same or different sensors
• Both support 8-bit unsigned, 16-bit signed/unsigned

Differences