Vector Extrapolation-Based Acceleration of Regularized Richardson Lucy Image Deblurring

Abstract

Confocal fluorescence microscopy has become an important tool in biological and medical sciences for imaging thin specimen, even living ones. Due to out-of-focus blurring and noise the acquired images are degraded and thus it is necessary to restore them. One of the most popular methods is an iterative Richardson-Lucy algorithm with total variation regularization. This algorithm while improving the image quality is converging slowly whereas with a constantly increasing amount of image data fast methods are required. In this paper, we present an accelerated version of the algorithm and investigate the achieved speed up. The acceleration method is based on a vector extrapolation technique and avoids a computational intensive evaluation of the underlying cost function. To evaluate the acceleration two synthetic test images are used. The accelerated algorithm reaches an acceptable result within 30% to 40% less computational time.