Abstract
Shape-from-focus (SFF) is a technique used to estimate the depth or 3D shape of an object from a sequence of images obtained at different focus settings. Most SFF algorithms compute the focus value of a pixel from a fixed square window surrounding the pixel. Choosing an optimum window size is an important issue in SFF; a small window size is vulnerable to pixel noise, whereas a larger widow size unnecessarily includes uncorrelated neighboring pixels. This paper demonstrates that window shape, which has yet to be effectively dealt with, is very important for improving the accuracy of focus measurements. In this paper, window size and shape are determined pixel-by-pixel using a semivariogram in order to maximally include correlated neighboring pixels. Experiments were then conducted on both synthetic and real objects, with the results confirming that the proposed technique improves the quality-of-focus measurement in comparison to previous methods based on a fixed square window.
► Surface radiance distribution using traditional 3D shape reconstruction methods. ► Optimal window size and shape selection using semivariogram. ► The focus measure based on the proposed adaptive window shape was compared with traditional focus measure based on fixed window shape. ► Accuracy improvement of focus measure. ► It produces better three dimensional shape.