Low Power Image Processing

This repo is a collection of different image processing algorithms that are able to be used on low powered devices in a reasonable amount of time. These can be used a image processing pipeline to improve the quality of the end result.

Recommended to process an image the normal route and then pick and chose what to use from here based on your device to create a seperate "improved" image

Devices tested on

• Pinephone Convergence Edition (Pine64)

Algorithms

• stacking

  • auto_stacking: Stack multiple images into a single image to reduce noise and improve quality

• super_resolution

  • opencv_super_resolution: Super resolution using OpenCV built-in methods via ESPCN or FSRCNN

• denoise

  • fastNlMeansDenoisingColored: Denoise using OpencV built-in methods
  • ircnn: Machine Learning denoiser
  • ffdnet: Machine Learning denoiser

Licensing

This repo is a combination of many other peoples work/ideas, as such there a multiple different licenses in use. Please check the individual folders for the license of the code in that folder, if there is no license file then the code is licensed under the license in the root of this repo.

• color:

  • image_adaptive_3dlut: Apache 2.0

• dehaze:

  • darktables: MIT

• denoise:

  • fast: MIT
  • ffdnet: MIT
  • ircnn: MIT

• stacking: MIT

• sharpen: MIT

• super_resolution:

  • opencv: Apache 2.0