Landsat 8 truecolor 8-bit composite creator

Uses a specific landsat image from landsat8 amazon cloud source to create an 8-bit true color composite (bands 4 3 and 2). Uses custom stretching to enhance visually the output image.

Download

Create root directory where everything will be located:

    mkdir landsattc

create a subdirectory to store the data and outputs:

    mkdir landsattc/data
    cd landsattc

copy the software from github repo:

    git init .
    git remote add origin https://github.com/jferencik/landsatrgb.git
    git pull origin master

Installation

1. install docker. Folow your OS instructions from https://docs.docker.com/install/

2. use the dockerfile to build an image

    docker build -t landsattc .

3. run tests (optionally) There are two test modules, for core functionality and for run functionality. The test modules needs the path to a folder where they can write. This path needs to be passed in when running the unittests like this:

    docker run --rm -v $(pwd):/landsat -e working_folder=/tmp landsattc python3 -m unittest discover -s /landsat

This wil take some time because the red image will be downloaded as part of the core tests.

Usage

1. check script command line options/capabilities

docker run --rm -v $(pwd):/landsat landsattc python3 /landsat/solution.py -h

usage: solution.py [-h] -wf WORKING_FOLDER
                   [-minperc MIN_PERCENTILES [MIN_PERCENTILES ...]]
                   [-maxperc MAX_PERCENTILES [MAX_PERCENTILES ...]]
                   [-pwr POWER_SCALES [POWER_SCALES ...]] [-no_data NO_DATA]
                   [-v VIEW]

optional arguments:
  -h, --help            show this help message and exit
  -wf WORKING_FOLDER, --working_folder WORKING_FOLDER
                        Folder where the Landsat8 images will be downloaded
                        (default: None)
  -minperc MIN_PERCENTILES [MIN_PERCENTILES ...], --min_percentiles MIN_PERCENTILES [MIN_PERCENTILES ...]
                        Per band (red, green blue), percentiles to use when
                        computing minimum value used to stretch the band
                        (default: (1, 1, 5))
  -maxperc MAX_PERCENTILES [MAX_PERCENTILES ...], --max_percentiles MAX_PERCENTILES [MAX_PERCENTILES ...]
                        Per band (red, green blue) percentiles to use when
                        computing maximum value used to stretch the band
                        (default: (99, 99, 99))
  -pwr POWER_SCALES [POWER_SCALES ...], --power_scales POWER_SCALES [POWER_SCALES ...]
                        Per band (red, green blue) power scales (default:
                        (1.1, 1.4, 2))
  -no_data NO_DATA, --no_data NO_DATA
                        Background value to be ignored during stretching,
                        common to all bands (default: 0)
  -v VIEW, --view VIEW  visualize the RGB image using pylab (default: False)

Create various true color composites

run with default settings or low contract true color:

docker run --rm -v $(pwd):/landsat landsattc python3 /landsat/solution.py -wf /landsat/data

Again this will take some time as the iages are downloaded to the data subfolder. Next time the downloaded data will be reused. When the script finishes, the map.jpg and tci.jpg should be written to the data subfolder.

    ll data

    total 180304
    drwxr-xr-x 2 root root     4096 Mar 11 11:04 ./
    drwxr-xr-x 6 root root     4096 Mar 11 11:00 ../
    -rw-r--r-- 1 root root 55306494 Mar 11 11:04 LC08_L1TP_107035_20190105_20190130_01_T1_B2.TIF
    -rw-r--r-- 1 root root 56562590 Mar 11 11:02 LC08_L1TP_107035_20190105_20190130_01_T1_B3.TIF
    -rw-r--r-- 1 root root 57928741 Mar 11 11:01 LC08_L1TP_107035_20190105_20190130_01_T1_B4.TIF
    -rw-r--r-- 1 root root    90576 Mar 11 11:04 map.jpg
    -rw-r--r-- 1 root root 14723053 Mar 11 11:04 tci.jpg

create a high contrast true color which I prefer:

docker run --rm -v $(pwd):/landsat landsattc python3 /landsat/solution.py -wf /landsat/data -pwr 2 2 2

Fell free to test various options like setting different min and max percentiles per band or using different per band power scales.

The script features two types of options that control how the bands are stretched. The percentiles, both min nad max aim at controlling the amount of red blue respectively green color the true color image will contain. For example to increase the amount of blue in the image use equal power scales and minperc and lower the maxperc for the blue band:

docker run --rm -v $(pwd):/landsat landsattc python3 /landsat/solution.py -wf /landsat/data -pwr 2 2 2 -maxperc 98 98 95 -minperc 1 1 1

this will create a blueish image

The second option is the power scaling. Use this option to control the luminosity of the true color image. For example to create a very light image and thus decrese the contrast between adjacent pixels use a higer value for power scaling like 5 for all bands:

    docker run --rm -v $(pwd):/landsat landsattc python3 /landsat/solution.py -wf /landsat/data -pwr 5 5 5