[INTERNAL] Color Normalized Spectral Sharpening in ENVI: which image bands are sharpened ?

Help Article Update 3

Image sharpening consists in merging a low-resolution image with a high-resolution image, with resampling to the high-resolution pixel size. CN spectral sharpening is an extension of the Color Normalized algorithm used to sharpen 3-band RGB images. The advantage of this algorithm is that it can be used to simultaneously sharpen any number of bands. For example, the CN spectral sharpening can be used to sharpen hyperspectral data using a multispectral image.

The higher spatial resolution and lower spectral resolution data is called the sharpening image. It is used to enhance the lower spatial resolution but higher spectral resolution bands of the input image.

The ENVI CN Spectral Sharpening tool sharpens input bands only if they fall within the spectral range of one of the sharpening image's bands. All other input bands are unchanged in the output. This Help Article explains how to determine which bands of the input image will be sharpened or not.

Note that the input image to sharpen must have wavelengths defined in its header. The sharpening image must have both wavelengths and full-width-half maximum (FWHM) values defined.

If there are not already recorded the wavelengths and FWHM values can be entered manually in the image ENVI header file following the steps below (in ENVI Classic):

-   Right click on the image filename in the available band list and select Edit Header

-   Select Edit Attributes then FWHM and/or Wavelengths

-   Record the requested values

The FWHM is the full spectral width at half maximum. It is illustrated on the image below:

   

The Spectral Range Δs of the sharpening bands is defined as follow

    λ-0.5*FWHM < Δs < λ+0.5*FWHM

With λ the band center wavelength and FWHM the full width-half maximum value


Exemple 1:

a. Low resolution input image to sharpen: it is a 4 band image

   wavelength : { 0.485000, 0.560000, 0.660000, 0.830000}

b. High resolution sharpening image: it is a panchromatic image

   wavelength : { 0.675000}
   FWHM : { 0.450000}

The Spectral range of the sharpening image is then equal to:

   0.675-0.5*0.45 < Δs < 0.675+0.5*0.45
   0.45 < Δs < 0.9

All the wavelengths of the input image are included in the spectral range of the sharpening image so all bands of the input image will be sharpened.


Exemple 2:

a. Low resolution input image to sharpen: it is a 4 band image

   Wavelength : { 0.485000, 0.560000, 0.660000, 0.830000}

b. High resolution sharpening image: it is a panchromatic image

   Wavelength : { 0.675000}
   FWHM : { 0.300000}

The Spectral range of the sharpening image is then equal to:

   0.675-0.5*0.3 < Δs < 0.675+0.5*0.3
   0.525 < Δs < 0.825

Only 2 bands (bands 2 and 3) of the input image are included in the spectral range of the sharpening image so only these 2 bands of will be sharpened. The 2 others (bands 1 and 4) will not change in the output.