MRAFUSION

Description

MRAFUSION creates high-resolution color images by fusing black-and-white panchromatic (PAN) and multispectral (MS) color images. Known as multi-resolution analysis (MRA), this image-fusion technique applies wavelet decomposition. MRAFUSION can fuse images acquired with the same or various sensors. Although other image-fusion techniques, such as Intensity, Hue, and Saturation (IHS), ImageLock Data Fusion, and the PANSHARP algorithm can also be used, MRAFUSION produces similar-to-superior sharpening results while preserving the spectral characteristics of the original images. Both 8-bit and 16-bit data types are supported.

Parameters

Name	Type	Caption	Length	Value range
FILI *	String	Name of input file	1 - 192	1 -
DBIC	Integer	Input multispectral channels	0 -
FILI_PAN *	String	Name of input panchromatic file	1 - 192	1 - 1
DBIC_PAN	Integer	Input panchromatic-image channel	0 - 1	Default: 1
SRCBGD	String	Source background value	0 - 192	Default: FILE
LEVELS	Integer	Decomposition levels	0 - 1	-1 - Default: 0
RADAPT	String	Adapt radiometric values to input MS images	0 - 3	YES | NO | MTF | LIN | ALL Default: YES
EDGESHAR	Float	Edge Sharpen	0 - 1	0.01 - 100.0 Default: 1.0
FILO *	String	Name of output file	1 - 192	1 -
DBOC	Integer	Output channel	0 -
FTYPE	String	Format of output file	0 - 4	Default: PIX
FOPTIONS	String	Options for output format	0 - 64
POPTION	String	Pyramid options	0 - 32	OFF | NEAREST | AVERAGE | MODE Default: AVERAGE
REPORT	String	Report mode	0 - 192	Quick links

* Required parameter

Parameter descriptions

FILI

The name of the input file that contains the low-resolution, multispectral (color) data to fuse with the high-resolution, panchromatic-image data.

DBIC

The multispectral channel or channels in the input file.

The channels you specify are fused with the high-resolution, panchromatic-image data.

You can specify any multispectral channel or channels, including those that are outside of the range specified as reference channels; however, using only the channels you specify as reference yields more accurate results.

Do not specify duplicate channels.

Ranges of channels or segments can be specified with negative values. For example, {1,-4,10} is internally expanded to {1,2,3,4,10}. When you are not specifying a range in this way, only 48 numbers can be specified explicitly.

FILI_PAN

The name of the input file that contains the high-resolution, panchromatic-image data.

DBIC_PAN

The channel that contains the high-resolution, panchromatic-image data from the input file.

You can specify only one channel. The default is channel 1.

SRCBGD

The pixels in the source images to consider as background (NoData).

The same rule is applied independently to both input images.

LEVELS

Specify the number of levels to be used for decomposing the image when applying wavelets.

If set to 0, the number of levels will be automatically determined by the function, using the ratio of PAN over MS. This is the default value.

RADAPT

Method to adapt the output spectral values to those of the input multispectral (MS) image.

Adapting the radiometry can better match to the original multispectral image. The two methods are linear regression based on a small sliding window and an estimate of the Modulation Transfer Function (MTF).

EDGESHAR

Amount of sharpening to image edges to apply. A value of 1.0 (default) means no edge sharpening. Values greater than 1.0 apply greater sharpening. Values greater than 2.0 are not recommended.

Applying edge sharpening makes image more visually appealing but may degrade radiometric similarity with the original multispectral (MS) data.

FILO

The name of the file to which to write the pansharpened, multispectral channels.

If you do not specify a file name, it is created with the extents that match the overlap of the input panchromatic file and the input file, by using the resolution of the input panchromatic file.

DBOC

The channel or channels to which to write the pansharpened channels.

If you do not specify a value for this parameter, the first n channels are used, where n is the number of selected input channels.

Do not specify duplicate channels.

This parameter is optional.

Ranges of channels or segments can be specified with negative values. For example, {1,-4,10} is internally expanded to {1,2,3,4,10}. When you are not specifying a range in this way, only 48 numbers can be specified explicitly.

FTYPE

The format of the output file.

The default value is PIX.

FOPTIONS

The options specific to the format to apply when creating the output file. With each, the default of no options is allowed (empty string).

Typically, the available options for a format include a compression scheme, format subtype, or other information.

POPTION

The resampling method to use to compute the overview levels.

You can select from the following options:

• OFF: Overviews are not produced. Although this option delays the browsing of an image, processing is complete quickly and a minimum amount of disk space is required.
• NEAREST: Applies a closest-neighbor resample. This is the default value.
• AVERAGE: Calculates the average value of all pixels that contribute to the overview pixel. This option is well suited to continuous-tone images and produces a better looking and more interpretable overview than NEAREST; however, do not use this option with classified, pseudocolor, or thematic images, because it introduces intermediate values not found in the original image.
• MODE: Applies the mode of the pixels that contribute to the overview pixel. Use this method with classified, pseudocolor, or thematic images, but not continuous-tone images.

Use guide

Processing time of building overviews with AVERAGE or MODE can be significantly slower than NEAREST. If you do not know the characteristics of an image, or if preparing the image overview quickly is important, use NEAREST.

REPORT

Specifies where to direct the generated report.

Available options are:

• TERM: generates a report on the terminal (default)
• DISK: generates a report on file "IMPRPT.LST"
• OFF: turns off report generation
• <filename>: appends a report to the specified file

Details

MRAFUSION applies an image-fusion algorithm to increase the resolution of multispectral (color) imagery by using a high-resolution, panchromatic (black and white) image.

You can use MRAFUSION with any data type (8-bit unsigned, 16-bit signed, 16-bit unsigned, 32 bit real) other than complex.

Typically the PAN (FILI_PAN) and MS (FILI) images are simultaneously acquired satellite or airphoto image. Landsat 7, Ikonos, Quickbird, and the SPOT series of satellite-borne sensors deliver PAN and MS image data that is acquired simultaneously. Using PAN and MS images from different sensors and/or from different times is possible but typically generates results of poorer quality. Images containing large amounts of cloud or other atmospheric effects may also impact the quality.

PAN and MS imagery must be perfectly coregistered before running MRAFUSION. Any misregistration will result in blurry imagery. If the output image quality is inferior or blurry, consider running geometric-correction preprocessing to improve the coregistration.

The ground-sample distances between the MS and PAN images should not exceed a four-to-one ratio for reasonable results. For example, Ikonos data with 4-meter MS and 1-meter PAN has a four-to-one ratio, which is suitable. However, Landsat 7 30-meter MS images have 50 times the ground-sample distance of QuickBird 0.61-meter PAN images, and attempting to sharpen the former from the latter will produce poor results.

Typically, the mean, standard deviation, and histogram shape for each MS image is only slightly modified by sharpening, provided the preceding guideline on the ratio of ground-sample distance is followed and the reference images cover closely the same wavelength range as the PAN image.

The EDGESHAR parameter allows extra sharpening to be applied to edges in the resulting imagery. This is useful if the imagery will be used for visual purposes, but can affect the radiometric similarity with the original lower resoution input MS imagery (FILI). An EDGESHAR value of 1.0 applies no extra sharpening (default) and a value of 1.5 or 2.0 apply higher amounts of edge sharpening. Values greater than 2 are possible but not recommended. Values less than 1 will blur the image.

If a pixel is NoData in either of the input images, it will be NoData in the output. Regardless of the value of the input NoData, the output NoData value will always be zero (0).

If an output file exists in the folder you specify, the image within the common bounding box of the panchromatic, multispectral and output channels will be pansharpened and saved to the output file. The sections of panchromatic or multispectral images that are not located within the common bounding box, but within the bounds of the output file, will be interpolated to the resolution of the output file using cubic convolution and then copied to the output file.

If an output file does not exist in the folder, a new file will be created with georeferenced extents equal to the common bounding box of the panchromatic and multispectral reference channels and pixel size equal to that of the panchromatic image.

Example

A QuickBird MS image in PCIDSK format, toronto_ms_demo.pix will be sharpened using the correpsonding QuickBird PAN image in a PCIDSK toronto_pan_demo.pix file. The sharpened images will be written to a new PCIDSK file, pansharp.pix.

The four MS images to be sharpened will also be used as the reference images.

EASI>FILI	 =	"toronto_ms_demo.pix"
EASI>DBIC	 =	1,2,3,4
EASI>FILI_PAN =	"toronto_pan_demo.pix"
EASI>DBIC_PAN =	1
EASI>SRCBGD	 =	"ANY, 0"	! Exclude zero-valued pixels in any input image
EASI>LEVELS	 =	0           ! Use system-computed number of levels
EASI>RADAPT	 =	"YES"       ! Apply radiometric adaptation
EASI>EDGESHAR =  1           ! No extra edge sharpening
EASI>FILO	 =	"pansharp.pix"
EASI>DBOC	 =	1,2,3,4
EASI>ftype	 =	"TIF"	    ! Create output file in GeoTIFF format
EASI>foptions =	"TILED512"	! Tile output file into tiles 512 x 512 pixels
EASI>POPTION  =	"AVERAGE"	! averaging resampling for pyramids
EASI>RUN MRAFUSION
            

Algorithm

The A Trous nonorthogonal wavelet transformation is used as the basis for pansharping images. The advantage of this type of transformation is that the reconstruction can perfectly recover the original image from decomposed ones regardless of how many decomposition levels have been performed. The transformation can also increase the spatial resolution of the MS image to that of the PAN while preserving the quality of the spectral information. Because of the multiple levels of decomposition, the transformation tends to provide pansharpened products that have less noise.

Acknowledgements

The automatic image-fusion algorithm was developed by PCI Geomatics.

References

M. Gonzalez-Audicanna, X. Otazy, O. Fors and A. Seco. (2005). "Comparison between Mallat's and the A Trous discrete wavelet transform based algorithm for the fusion of multispectral and panchromatic images". In International Journal of Remote Sensing, Vol. 26, No. 3, pp 595-614.