README.md

MULTISPECTRAL IMAGE PAN-SHARENING

There is one source file in this repository: a single Python (.py) 
source file. The Python file pansharpen.py is a stand-alone program that 
is to be run on the UNIX/Linux command line environent. The first required 
argument should be the filename of a 1-band panchromatic Geotiff image file
(.tif extension). The second argument should be a string filename of a 3 or 
4-band multispectral geotiff image file (RGB,NIR bands, in that order).
The Python program does a number of tasks. 
   
First, it uses GDAL tools to 
resample the multispectral Geotiff image file to the same higher dimensions 
as the panchromatic image Geotiff file using bicubic interpolation. 
This file is then written to disk. For this resampled multispectral Geotiff image
file and the panchromatic Geotiff image file, the following pan-sharpening
algorithms are applied:
   
  (1) Brovey, 
  (2) Principal Component Analysis (PCA),
  (3) FIHS (Fast Intensity Hue Saturation),
  (4) Wavelet

DISCLAIMER:

  This program is heavily dependent on how much memory you have. It stores the data
  into large arrays for the pan-sharpening computations. This is not efficient.
  In the future, a tiling procedure may be implemented (time allowing). For now, 
  try to use images no larger than 5000x5000 pixels or so. Depending on your system.
  
  You may need to "tile up" or "slice and dice" your panchromatic and multispectral
  imagery before using this program on each tile individually.

DESCRIPTION:

This program performs pansharpening of satellite imagery. It is meant to be 
run on the command-line on UNIX-like operating systems. The two primary inputs
are (1) a 3 or 4 band multispectral geotiff containing the red, green, blue, 
and NIR bands (NIR is optional) and (2) a 1-band geotiff containing higher-
resolution greyscale panchromatic image data. It is assumed that both of 
these two Geotiff inputs are "clipped" to the same rectangular geographic 
bounding-box. Four methods of pan-sharpening are used: Brovey, Fast Intensity
Hue Saturation (FIHS), Wavelet, and Principal Component Analysis (PCA).

INSTALLATION:

To use this software, it is expected you will have docker (https://www.docker.com)
installed.

$ git clone https://github.com/gerasimosmichalitsianos/pansharpen
$ cd pansharpen/
$ docker build -t pansharpen .

USAGE:

To use this code, pass in the filenames of the panchromatic and multispectral 
image (e.g. Geotiffs) files e.g.

$ ls /home/username/DATA
PAN.TIF  RGB_NIR.TIF  RGB.TIF
$ DIR=/home/username/DATA
$ docker run -v $DIR:$DIR pansharpen --multispectral $DIR/RGB.TIF --panchromatic $DIR/PAN.TIF

Options: 
  --version,       -v : display version info.
  --usage,         -u : display this usage messsage
  --panchromatic,  -p : pass in name of 1-band Geotiff holding 1-band panchromatic Geotiff image (high resolution, required)
  --multispectral, -m : pass in name of 3 or 4 band multispectral Geotiff image file (low-resolution, required)

EXAMPLE USAGE:

$ multispectralGeotiff=LC08_L1TP_185033_20170712_20170726_01_T1_MULTI_TOA_3BAND.TIF
$ panchromaticGeotiff=LC08_L1TP_185033_20170712_20170726_01_T1_B8_TOA.TIF
$ docker run -v $(pwd):$(pwd) --panchromatic $(pwd)/$panchromaticGeotiff --multispectral $(pwd)/$multispectralGeotiff

OUTPUTS: 
  When the program is complete, there should be 4 new Geotif image files: 
    (1) a 3 or 4 band multispectral Geotiff image created using Brovey pan-sharpening
    (2) a 3 or 4 band multispectral Geotiff image created using FIHS pan-sharpening
    (3) a 3 or 4 band multispectral Geotiff image created using Wavlet pan-sharpening
    (4) a 3 or 4 band multispectral Geotiff image created using PCA pan-sharpening
      
  NOTE: These outputs should be in the same directory as the input files passed-in via command-line.

PYTHON VERSION:

Supports Python 3.x

Sample Outputs

The below sample images show the results of this algorithm using Landsat 8 imagery over Skala (Σκάλα), 
Greece. Skala is a small Greek town found on Greece's island of Kefalonia in western Greece.

Left: original panchromatic image.
Center: pan-sharpened RGB image using Brovey technique.
Right: pan-sharpened RGB image using FIHS (Fast Intensity Hue Saturation) technique.

Left: original RGB low-resolution image.
Center: pan-sharpened RGB image using Wavelet technique.
Right: pan-sharpened RGB image using PCA (Principal Component Analysis) technique.

@author:

Gerasimos Michalitsianos
[email protected]
8 February 2021