π-xel › Image processing › Adaptive Focusing
What it does?
'Adaptive Focusing' is an image filtering algorithm which produce smoother images. Those are generally compressed more eficiently
by standard algorithms (e.g. JPEG, WEBP or the modern Lepton) - between 5% and 25% depending on the complexity of the original image and the desired output quality.
How it works?
'Adaptive Focusing' is basically using classical image filtering algorithms like blur, pixelate, posterize (which normaly produce information loose). Applying such filters normaly produce more compressable but low quality images and the visual degradation is usualy very unplesant and unaceptable.
However, we are using sophisticated machine vision algorithms to apply the filtering process only on some specific areas of the images, minimazing the visual degradation of the original image.
More comparative results
Adaptive Focusing filter was applied on a well known test images dataset (KODAK test images
) then encoded with JPEG and WEBM. Both cases shows similar
size reduction relative to the filter quality factor.
To test it on your own, download the precompiled Linux binary
It woks with PPM images, so use imagemagik
# convert original.jpg original.ppm
# ./pixel -i original.ppm -o original_out.ppm -b 1 -q 3
# convert original_out.ppm -quality 95 -sampling-factor 4:2:0 original_out.jpg
# convert original.ppm -quality 95 -sampling-factor 4:2:0 original_reencoded.jpg
More of our tools and ideas
Adaptive Focusing for video processing. Make your videos 5 to 10% smaller with unnoticible quality loose.
A lightweight general purpose compression tool suitable for IoT.
Check source code on GitHub
To be available soon.
Our roadmaps includes:
- Linux and Android distributions
- pure software processing
- hardware accelerated processing using OpenGL-ES support
- integration with major open source media frameworks