A Plug-and-Play Priors Framework for Hyperspectral Unmixing

Spectral unmixing is a widely used technique in hyperspectral image processing and analysis. It aims to separate mixed pixels into the component materials their corresponding abundances. Early solutions spectral are performed independently on each pixel. Nowadays, investigating proper priors problem has been popular as it can significantly enhance performance. However, nontrivial handcraft powerful regularizer, complex regularizers may introduce extra difficulties solving optimization problems which they involved. To address this issue, we present plug-and-play (PnP) framework for unmixing. More specifically, use alternating direction method of multipliers (ADMM) decompose two iterative subproblems. One regular depending forward model, other proximity operator related prior model be regarded an denoising problem. Our flexible extendable allows wide range denoisers replace models avoids handcrafting regularizers. Experiments conducted both synthetic data real airborne illustrate superiority proposed strategy compared with state-of-the-art methods.