A new localized and computationally efficient approach is presented for shift/space-variant image restoration. Unlike conventional approaches, it models shift-variant blurring in a completely local form based on the recently proposed Rao transform (RT). RT facilitates almost exact inversion of the blurring process locally and permits very fine-grain parallel implementation. The new approach naturally exploits the spatial locality of blurring kernels and smoothness of underlying focused images. It formulates the deblurring problem in terms of local parameters that are less correlated than raw image data. It is a fundamental advance that is general and not limited to any specific form of the blurring kernel such as a Gaussian. It has significant theoretical and computational advantages in comparison with conventional approaches such as those based on singular value decomposition of blurring kernel matrices. Experimental results are presented for both synthetic and real image data. This approach is also relevant to solving integral equations.

A new approach is presented for 3D shape recovery of local planar surface patches from two shift/space-variant defocused images. It is based on a recently proposed technique for inverting the shift-variant blurring in a camera system. It is completely localized, applicable to general point spread functions, and facilitates fine-grain parallel implementation and computationally efficient recovery of 3D shape. The shape of a visible object at each pixel is specified by the distance and slopes of a local planar surface patch that approximates the object surface at that pixel. The theory underlying the new approach and experimental results are presented. Results of both simulation and real experiments indicate that the new approach is effective and useful in practical applications.

Automatic reconstruction of a complete 3D model of a complex object is presented. The complete 3D model is reconstructed by integrating two 3D models which are reconstructed from different poses of the object. For each pose of the object, a 3D model is reconstructed by combining stereo image analysis, shape from silhouettes, and a volumetric integration technique. Stereo image analysis and shape from silhouettes techniques complement each other to reconstruct an accurate and noise-resistant 3D model. For a reliable volumetric integration of multiple partial shapes, a voxel coding technique is introduced. The voxel coding technique facilitates a selection of consistent partial shapes for shape integration. In order to reconstruct all visible surfaces of a complex object with concavities and holes, two 3D models from different poses of the object are reconstructed and integrated to obtain the complete 3D model. A voxel coding technique is again used during pose integration. Experimental results on a real object demonstrate that our approach has advantages and is effective.

An automatic 3D model reconstruction technique is presented to acquire complete 3D models of real objects. The technique is based on novel approaches to pose estimation and integration. Two different poses of an object are used because a single pose often hides some surfaces from a range sensor. The presence of hidden surfaces makes the 3D model reconstructed from any single pose a partial model. Two such partial 3D models are reconstructed for two different poses of the object using a multi-view 3D modeling technique. The two partial 3D models are then registered. Coarse registration is facilitated by a novel pose estimation technique between two models. The pose is estimated by matching a stable tangent plane (STP) of each pose model with the base tangent plane (BTP) which is invariant for a vision system. The partial models are then integrated to a complete 3D model based on voxel classification defined in multi-view integration. Texture mapping is done to obtain a photo-realistic reconstruction of the object.

A desktop vision system is presented for complete 3D model acquisition. It is fast, low-cost, and accurate. Partial 3D shapes and texture information are acquired from multiple viewing directions using rotational stereo and shape from focus (SFF). The resulting range images are registered to a common coordinate system and a surface representation is created for each range image. The resulting surfaces are integrated using a new algorithm named Region-of-Construction. Unlike previous approaches, the Region-of-Construction algorithm directly exploits the structure of the raw range images. The algorithm determines regions in range images corresponding to non-redundant surfaces which can be stitched along the boundaries to construct the complete 3D surface model. The algorithm is computationally efficient and less sensitive to registration error. It also has the ability to construct complete 3D models of complex objects with holes. A photo realistic 3D model is obtained by mapping texture information onto the complete surface model representing 3D shape. Experimental results for several real objects are presented.

Two new methods are presented for recovering the focused image of an object from only two blurred images recorded with different camera parameter settings.The camera parameters include lens position, focal length, and aperture diameter.First a blur parameter is estimated using one of our two recently proposed depth-fromdefocus methods.Then one of the two blurred images is deconvolved to recover the focused image.The first method is based on a recently proposed Spatial Domain Convolution/Deconvolution Transform.This method requires only the knowledge of of the camera's point spread function (PSF).It does not require information about the actual form of the camera's PSF.The second method, in contrast to the first, requires full knowledge of the form of the PSF.As part of the second method, we present a calibration procedure for estimating the camera's PSF for different values of the blur parameter .In the second method, the focused image is obtained through deconvolution in the Fourier Domain using the Wiener filter.For both methods, results of experiments on actual defocused images recorded by a CCD camera are given.The first method requires much less computation than the second method.The first method gives satisfactory results for up to medium levels of blur and the second method gives good results for up to relatively high levels of blur.

Two new methods are presented for recovering the focused image of an object from only two blurred images recorded with different camera parameter settings. First a blur parameter /spl sigma/ is estimated using one of our two recently proposed depth-from-defocus methods. Then one of the two blurred images is deconvolved to recover the focused image. The first method requires only the knowledge of /spl sigma/ of the camera's point spread function (PSF). It is based on a new spatial domain convolution/deconvolution transform. The second method requires full knowledge of the form of the PSF and is estimated using step edges. The focused image is obtained through deconvolution in the Fourier Domain using the Wiener filter. For both methods, results of experiments on actual defocused images recorded by a CCD camera are presented. The first method requires less computation and gives satisfactory results for up to medium levels of blur. The second method gives good results for up to relatively high levels of blur.>

The application of a new spatial-domain convolution/deconvolution transform (S transform) for determining the distance of objects using image defocus is described. The method, known as STMAP, involves simple local operations on only two images taken with different aperture diameters and can be easily implemented in parallel. Both images can be arbitrarily blurred, and neither of them needs to be a focused image taken with a pin-hole camera. STMAP is implemented on an actual camera system named the Stonybrook Passive Autofocusing and Ranging Camera System (SPARCS). Experimental results on real-world planar objects are presented. The results indicate that STMAP is useful in practical applications. The utility of the method is demonstrated for rapid autofocusing of electronic cameras. STMAP is computationally more efficient than other depth-from-focus methods. The results are comparable to a Fourier-transform-based approach.>

A method for determining depth (range) from image defocus and rapid autofocusing of a camera named DFD1F, is presented. It requires only two images in theory, but three images in the present implementation. DFD1F is based on computing only one-dimensional Fourier coefficients as opposed to two-dimensional Fourier coefficients for a related prior method, thus providing not only computational advantage but also robustness in practical applications. DFD1F is independent of the form of the modulation transfer function of the camera. DFD1F has been successfully implemented and tested on an actual camera system named SPARCS. SPARCS can determine the distance of an object placed in front of it at any distance in the range of 0.5 m to infinity, and can successfully focus the object by moving the lens with a root-mean-square error of less than 6% in terms of lens position.>

A new method is described for recovering the distance of objects in a scene from images formed by lenses. The recovery is based on measuring the change in the scene's image due to a known change in the three intrinsic camera parameters: (i) distance between the lens and the image detector, (ii) focal length of the lens, and (iii) diameter of the lens aperture. The method is parallel involving simple local computations. In comparison with stereo vision and structure-frommotion methods, the correspondence problem does not arise. This method for depth-map recovery may also be used for (i) obtaining focused images (i.e. images having large depth of field) from two images having finite depth of field, and (ii) rapid autofocusing of computer controlled video cameras. 1. Introduction Here we describe a new passive ranging method which in principle is fast and involves relatively weak assumptions that are generally valid. The method is basically a generalized version of the `depth-from-focu...