top of page
Writer's pictureGeoinformers

Image Resolution Enhancement using spectral sharpening


The spectral sharpening methods are very much useful for image resolution enhancement. It has been observed that few satellite sensors like LISS IV provide 5.8 m resolution but if we perform spectral sharpening on the imagery, we can enhance the image resolution considerably and this enhanced resolution image will help greatly in extracting features of interest.

To perform spectral sharpening, The source images must be georeferenced to a standard map projection. If Low Resolution image and High Resolution image have different projections, the low resolution image is needed to be transformed into high resolution image's projection system.

According to Harris geospatial solutions, "Pan-sharpening algorithms are used to sharpen multispectral data using high spatial resolution panchromatic data. An underlying assumption of these algorithms is that you can accurately estimate what the panchromatic data would look like using lower spatial resolution multispectral data."

In the ENVI image processing software, one can also write a script in IDL to perform pan sharpening using the 'ENVIGramSchmidtPanSharpeningTask' routine.

Example code of the task is given on this link.

Here is a case study of Resourcesat-2 satellite's LISS-IV sensor image. Spectral sharpening operations were performed on this imagery.

The LISS-IV image before spectral sharpening is shown below-

Fig-1


The most commonly used method of spectral sharpening are-

HSV Sharpening

HSV (Hue, Saturation, Value) sharpening is used to transform an image from RGB to HSV color space. It converts the low resolution input image into high resolution image. It requires input as low resolution image and one high resolution band. It re-samples the hue and saturation bands to high resolution pixel size using any of three re-sampling techniques (nearest neighbor, bilinear, and cubic convolution) available in ENVI image processing software.

The results of HSV sharpening on LISS-IV image -

Overall hue and saturation are improved and the fields having more water content is depicted in greenish tone.

Fig-2


Gram Schmidt Spectral Sharpening

Use Gram-Schmidt (GS) Spectral Sharpening to sharpen multispectral data using high spatial resolution data. If both data sets are georeferenced, ENVI Classic additionally co-registers them on-the-fly.

Steps involved in GS spectral sharpening-

  • Compute a simulated low resolution Pan band as a linear combination of the n Multi-spectral bands.

  • Performing a Gram-Schmidt transformation on the simulated panchromatic band and the spectral bands, using the simulated panchromatic band as the first band.

  • Replace the low resolution simulated Pan band by the gain and bias adjusted high resolution Pan band. Upsample all Multi-spectral bands accordingly.

  • Reverse the forward Gram-Schmidt transform using the same transform coefficients, but on the high resolution bands. The result of this backward Gram Schmidt transform is the pan-sharpened image in high resolution.

Why Gram Schmidt Spectral Sharpening is better?

The Gram-Schmidt and PC spectral sharpening tools both create pan-sharpened images, but using different techniques. Generally speaking, the Gram-Schmidt method is more accurate than the PC method and is recommended for most applications. Gram-Schmidt is typically more accurate because it uses the spectral response function of a given sensor to estimate what the panchromatic data look like. If you display a Gram-Schmidt pan-sharpened image and a PC pan-sharpened image, the visual differences are very subtle. The differences are in the spectral information; compare a Z Profile of the original image with that of the pan-sharpened image to see the differences in spectral information, or calculate a covariance matrix for both images. The effect of pan sharpening is best revealed in images with homogenous surface features (Harrisgeospatial, ENVI Help)

Fig-3


PC Spectral Sharpening

Four steps are involved in PC (principal component) spectral sharpening technique-

  • Principal component transformation on all the bands of multispectral data.

  • Replacing PC band 1 with the high resolution band and scaling the high resolution band to match the PC band 1, so no distortion of the spectral information occurs. The PC spectral sharpening method assumes that the first PC band is a good estimate of the panchromatic data (Harrisgeospatial, ENVI Help).

  • Applying inverse PC transform

  • Finally resampling of multispectral data into high resoultion image pixel size using an appropriate algorithm selected by user.

Fig-4


CN Spectral Sharpening

Its algorithm is also referred as Energy Subdivision Transform. The beauty of CN (Colour Normalized) spectral sharpening is that one can Using CN Spectral Sharpening simultaneously sharpen any number of bands and retain the input image’s original data type and dynamic range. for instance, we can sharpen hyperspectral image band using a multispectral image.

The content help of ENVI software provides a comprehensive information about CN sharpening and says that the spectral range of the sharpening bands are defined by the band center wavelength and full width-half maximum (FWHM) value, both obtained from the sharpening image’s (high resolution image) header file. The input image’s bands are grouped into spectral segments defined by the spectral range of the sharpening bands. The corresponding band segments are processed together in the following manner. Each input band is multiplied by the sharpening band, then normalized by dividing by the sum of the input bands in the segment

Fig-5


Comparative illustration of images



Reference:

  • T. Maurer, 2013, HOW TO PAN-SHARPEN IMAGES USING THE GRAM-SCHMIDT PAN-SHARPEN METHOD – A RECIPE, International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-1/W1, ISPRS Hannover Workshop 2013, 21 – 24 May 2013, Hannover, Germany

  • Harrisgeospatial, Using Image Sharpening, ENVI help

Author

Gopal Krishna

PhD-Geoinformatics and Remote Sensing


492 views0 comments

header.all-comments


Header-edited.jpg
bottom of page