This repository contains the code related to the 3CPS supercompression proposal, which is introduced in Hristina Hristova, Gwendal Simon, Stefano Petrangeli, and Vishy Swaminathan. ''3CPS: A Novel Supercompression for the Delivery of 3D Object Textures, in ACM Multimedia Systems (MMSys) Conference, 2020.
It performs a three-pass compression based on DXT1. The re-compression step, based on the Maximum Difference Cut, is implemented in the NVIDIA Texture Tools (repository: https://github.com/castano/nvidia-texture-tools).
A primary feature of NVIDIA Texture Tools is the DXT compression. Our solution contributes to the nvtt library, optimizing the CompressorDXT1 functionality (including functions for performing DXT1 texture compression).
To install and run our 3CPS solution, you can use the support for building Docker image and running encoding experiments in a Docker container. The folder experimental_data/textures/ contains uncompressed examplar textures.
$ docker build -t 3cps-nvidia-texture-tools:latest .
Always run the software from the root folder of the git repository.
To run the software for all examplar textures in experimental_data/textures/ folder:
$ docker run -ti -v $(pwd)/experimental_data:/nvidia-texture-tools/data --rm 3cps-nvidia-texture-tools
To run the software for a particular texture t.jpg from the experimental_data/textures/ folder:
$ docker run -ti -v $(pwd)/experimental_data:/nvidia-texture-tools/data --rm 3cps-nvidia-texture-tools t.jpg
Moreover, to compare the re-compression step from our 3CPS solution to the original DXT1 compression, build the Docker image in the folder base-nvidia-texture-tools/. The build will clone the original nvidia-texture-tools repository and install the software. Then, run the Docker container for all examplar textures or a particular texture.
The compressed images are stored in folders experimental_data/3cps/ (for the 3CPS supercompression method) and experimental_data/dxt1/ (for the DXT1 method from the NVIDIA Texture Tools).