Analysis and application of autofocusing and three-dimensional shape recovery techniques based on image focus and defocus

Autofocusing and three-dimensional (3D) shape recovery techniques based on image focus and defocus are analyzed and their practical applications are demonstrated. Focus measures based on summing the absolute values of image derivatives are used by many researchers in the past. We first investigated the unsoundness of those focus measures. We also argued that energy of the Laplacian of the image is a good focus measure and is recommended for use in practical applications. Important properties of the Laplacian focus measure are investigated. Application of the Laplacian focus measure to 3D microscopy is demonstrated.; The optimal focus measure for a noisy camera in passive search based autofocusing (AF) and depth-from-focus (DFF) applications depends not only the camera characteristics but also the image of the object being focused or ranged. In the early stage of this research, a new metric named Autofocusing Uncertainty Measure (AUM) was defined which is useful in selecting the most accurate focus measure from a given set of focus measures. AUM is a metric for comparing the noise sensitivity of different focus measures. In the later stage of this research, an improved metric named Autofocusing Root-Mean-Square Error (ARMS error) was defined. Explicit expressions have been derived for both AUM and ARMS error, and the two metrics are shown to be related by a monotonic expression. The theories are verified by experiments as well as computer simulations.; Another ranging method, depth-from-defocus (DFD) using the Spatial Domain Convolution/Deconvolution Transform Method (STM), is an useful technique for autofocusing and 3D shape recovery. The noise sensitivity analysis of STM is investigated. A theoretical treatment of this problem provides the accuracy check of STM in the presence of noise which has been done only under experimental observation. The derived theoretical results and supporting experimental results are presented.; Finally, the integration of DFF and DFD methods is developed for fast and accurate 3D shape recovery. This has been demonstrated successfully with experiments on a prototype camera system.