K-Means Image Segmentation

K-Means image segmentation (feature extraction) that just works. Lightweight and low footprint C++ implementation.

Build

1. Initialize and update the Git submodules: git submodule init && git submodule update.
2. Compile with make. Can also compile for ARM architecture with make TARGET=arm.

This program was specifically created for the European Space Agency's OPS-SAT Spacecraft to enable in-orbit cloud detection and feature extraction with pictures acquired by the on-board camera. As such, the program's stdout is specific to OPS-SAT's needs. However, this specificity can be disabled by setting TARGET_BUILD_OPSSAT to 0 in main.cpp:

/* indicate if we are building for the OPS-SAT spacecraft or not */
#define TARGET_BUILD_OPSSAT         0

Alternatively, developers are encouraged to fork this repo and remove OPS-SAT related logic.

Usage

Easy to use!

Help

$ ./image_segmentation -?
image_segmentation [options] ...
  --input    / -i       the file path of the input image
  --write    / -w       the write mode of the output image (optional)
        0 - do not write a new image (equivalent to not specifying --write)
        1 - write a new image as a new file
        2 - write a new image that overwrites the input image file
        3 - same as option 2 but backs up the original input image
  --kvalue   / -k       the k-means k value (2 to 11)
  --palette  / -p       the palette to use for segmentation (optional)
  --help     / -?       this information

Palette options (case sensitive):

• Random
• RdYlBu
• PiYG
• Paired
• Set3
• BW

Refer to colorbrewer2.org for palette colors. Supported range for k values are 2 to 11 for palettes except BW (k=2) and Random (k >= 2 with no ceiling).

Example

$./image_segmentation -i samples/img_msec_1607377664281_2_thumbnail.jpeg -w 1 -k 7 -p Set3