Automatic estimation of point-spread-function for deconvoluting out-of-focus optical coherence tomographic images using information entropy-based approach

This paper proposes an automatic point spread function (PSF) estimation method to de-blur out-of-focus optical coherence tomography (OCT) images. The method utilizes Richardson-Lucy deconvolution algorithm to deconvolve noisy defocused images with a family of Gaussian PSFs with different beam spot sizes. Then, the best beam spot size is automatically estimated based on the discontinuity of information entropy of recovered images. Therefore, it is not required a prior knowledge of the parameters or PSF of OCT system for de-convoluting image. The model does not account for the diffraction and the coherent scattering of light by the sample. A series of experiments are performed on digital phantoms, a custom-built phantom doped with microspheres, fresh onion as well as the human fingertip in vivo to show the performance of the proposed method. The method may also be useful in combining with other deconvolution algorithms for PSF estimation and image recovery.