All denoise options in Blender are based on different research efforts to speed up path tracing.
is an implementation exclusively for Cycles by Lucas Stockner during Summer of Code 2016 combining different papers about denoising around 2015:
...this proposal is about having a denoiser right in Cycles - where all the additional information (like feature passes and variance info) is available and can be used to produce results that are far better than general image denoising, often allowing to cut render times by 75% or more.
is a research effort on top of OptiX a GPU library which requires a graphics card supporting CUDA. The Paper was published at SIGGRAPH 2017 and can be found here:
It uses GPU-accelerated artificial intelligence to dramatically reduce the time to render a high fidelity image that is visually noiseless.
is a library build on top of DNNL (a CPU libary for deep learning applications) published in 2018:
Intel Open Image Denoise internally builds on top of Intel® Deep Neural Network Library (DNNL), and automatically exploits modern instruction sets like Intel SSE4, AVX2, and AVX-512 to achieve high denoising performance. A CPU with support for at least SSE4.1 is required to run Intel Open Image Denoise.
Related: Which denoiser is better?