Sparse Shape From Focus: Joint Estimation of 3D shape and Focus Image of 3D object

Shape-from-focus (SFF) is a traditional method of estimating the 3D structure of microscopic objects by measuring the degree of focus in a stack of space-variantly blurred images captured with relative motion between camera and the specimen. 3D shape of an object can be obtained if we have a perfectly focused image of the 3D object and a set of blurred images.Previous works in SFF have used a large number of images of the captured stack to reconstruct the focused image using a focus measure operator. The estimated focus image using this method is not perfect. Since the blurred images from which it is reconstructed could be affected by noise. So we proposed an alternating minimization method which uses sparsity in shape from focus technique to estimate the focus image and shape of 3D object.

(a) Shape from focus setup

(b) Formation of defocus blur