Fully Automatic Blind Color Deconvolution of Histological Images Using Super Gaussians

In digital pathology blind color deconvolution techniques separate multi-stained images into single stained bands. These band images are then used for image analysis and classification purposes. This paper proposes the use of Super Gaussian priors for each stain band together with the similarity to a given reference matrix for the color vectors. Variational inference and an evidence lower bound are then utilized to automatically estimate the latent variables and model parameters. The proposed methodology is tested on real images and compared to classical and state-of-the-art methods for histopathological blind image color deconvolution. Its use as a preprocessing step in prostate cancer classification is also analysed.