Showing papers on "Image registration published in 1994"

Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition

Abstract: Presents a new method to match a 2D image to a translated, rotated and scaled reference image. The approach consists of two steps: the calculation of a Fourier-Mellin invariant (FMI) descriptor for each image to be matched, and the matching of the FMI descriptors. The FMI descriptor is translation invariant, and represents rotation and scaling as translations in parameter space. The matching of the FMI descriptors is achieved using symmetric phase-only matched filtering (SPOMF). The performance of the FMI-SPOMF algorithm is the same or similar to that of phase-only matched filtering when dealing with image translations. The significant advantage of the new technique is its capability to match rotated and scaled images accurately and efficiently. The innovation is the application of SPOMF to the FMI descriptors, which guarantees high discriminating power and excellent robustness in the presence of noise. This paper describes the principle of the new method and its discrete implementation for either image detection problems or image registration problems. Practical results are presented for various applications in medical imaging, remote sensing, fingerprint recognition and multiobject identification. >

An Interactive Technique for Three-Dimensional Image Registration: Validation for PET, SPECT, MRI and CT Brain Studies

Abstract: 0.32to 2.22 mm (a.d.)and of rotationangles from0.32 to 1.70 degrees. ft was always much smaller than the point-spreed full-widthhalf maximum of the device with the lower resolution. The accuracyof coregistration was examinedusingtwo arbi tredlymiep@ced imagesets.Inter1nd@dUal andintraindMdual variance were similar, which suggested that the influence of subjectivity was not significant. Average displacements after coregistration were 0.43 and 0.29 mm or less for PET and MRI data, respectively, which indicated the absence of a systematic bias. Conclusion: The results indicate the high reproducibility and accuracy ofthus three-dimenalonal coreg@trationtethnk@ue, which is comparable or superior to those of automated tech niquesand methodsbased on externalartificiallandmarks.

A patient‐to‐computed‐tomography image registration method based on digitally reconstructed radiographs

Abstract: An automatic method for the accurate registration of computed tomography (CT) data with two camera-calibrated radiographs is presented. The registration is based on the skull as visualized both in the plain radiographs and in radiographs digitally reconstructed from CT. A reference coordinate system is established based on the radiographic projection parameters obtained using an angiographic stereotactic localizer. The CT-derived reconstructed radiographs are aligned iteratively at multiple resolutions until a best match is found by adjusting the position and orientation of the CT data set relative to the reference coordinate system. The results of experiments with a skull phantom performed under stereotactic control which show that reliable registration is possible with an accuracy better than 1 mm are presented. Possible applications include intraoperative patient-to-CT frameless registration and registration of radiographic data with frameless CT for depth electroencephalogram electrode position confirmation.

A moment-based approach to registration of images with affine geometric distortion

Abstract: A new method is described for automatic control point selection and matching. First, reference and sensed images are segmented and closed-boundary regions are extracted. Each region is represented by a set of affine-invariant moment-based features. Correspondence between the regions is then established by a two-stage matching algorithm that works both in the feature space and in the image space. Centers of gravity of corresponding regions are used as control points. A practical use of the proposed method is demonstrated by registration of SPOT and Landsat TM images. It is shown that the authors' method can produce subpixel registration accuracy. >

An efficient method for dynamic magnetic resonance imaging

Abstract: Many magnetic resonance imaging applications require the acquisition of a time series of images. In conventional Fourier transform based imaging methods, each of these images is acquired independently so that the temporal resolution possible is limited by the number of spatial encodings (or data points in the Fourier space) collected, or one has to sacrifice spatial resolution for temporal resolution. In this paper, a generalized series based imaging technique is proposed to address this problem. This technique makes use of the fact that, in most time-sequential imaging problems, the high-resolution image morphology does not change from one image to another, and it improves imaging efficiency (and temporal resolution) over the conventional Fourier imaging methods by eliminating the repeated encodings of this stationary information. Additional advantages of the proposed imaging technique include a reduced number of radio-frequency (RF) pulses for data collection, and thus lower RF power deposition. This method should prove useful for a variety of dynamic imaging applications, including dynamic studies of contrast agents and functional brain imaging.

Voxel similarity measures for automated image registration

Abstract: We present the concept of the feature space sequence: 2D distributions of voxel features of two images generated at registration and a sequence of misregistrations. We provide an explanation of the structure seen in these images. Feature space sequences have been generated for a pair of MR image volumes identical apart from the addition of Gaussian noise to one, MR image volumes with and without Gadolinium enhancement, MR and PET-FDG image volumes and MR and CT image volumes, all of the head. The structure seen in the feature space sequences was used to devise two new measures of similarity which in turn were used to produce plots of cost versus misregistration for the 6 degrees of freedom of rigid body motion. One of these, the third order moment of the feature space histogram, was used to register the MR image volumes with and without Gadolinium enhancement. These techniques have the potential for registration accuracy to within a small fraction of a voxel or resolution element and therefore interpolation errors in image transformation can be the dominant source of error in subtracted images. We present a method for removing these errors using sinc interpolation and show how interpolation errors can be reduced by over two orders of magnitude.

Three-dimensional registration using range and intensity information

Abstract: The determination of relative pose between two range images, also called registration, is a ubiquitous problem in computer vision, for geometric model building as well as dimensional inspection. The method presented in this paper takes advantage of the ability of many active optical range sensors to record intensity or even color in addition to the range information. This information is used to improve the registration procedure by constraining potential matches between pairs of points based on a similarity measure derived from the intensity information. One difficulty in using the intensity information is its dependence on the measuring conditions such as distance and orientation. The intensity or color information must first be converted into a viewpoint-independent feature. This can be achieved by inverting an illumination model, by differential feature measurements or by simple clustering. Following that step, a robust iterative closest point method is then used to perform the pose determination. Using the intensity can help to speed up convergence or, in cases of remaining degrees of freedom (e.g. on images of a sphere), to additionally constrain the match. The paper will describe the algorithmic framework and provide examples using range-and-color images.

Computerized detection of masses in digital mammograms: automated alignment of breast images and its effect on bilateral-subtraction technique.

Abstract: An automated technique for the alignment of right and left breast images has been developed for use in the computerized analysis of bilateral breast images. In this technique, the breast region is first identified in each digital mammogram by use of histogram analysis and morphological filtering operations. The anterior portions of the tracked breast border and computer-identified nipple positions are selected as landmarks for use in image registration. The paired right and left breast images, either from mediolateral oblique or craniocaudal views, are then registered relative to each other by use of a least-squares matching method. This automated alignment technique has been applied to our computerized detection scheme that employs a nonlinear bilateral-subtraction method for the initial identification of possible masses. The effectiveness of using bilateral subtraction in identifying asymmetries between corresponding right and left breast images is examined by comparing detection performances obtained with various computer-simulated misalignments of 40 pairs of clinical mammograms. Based on free-response receiver operating characteristic and regression analyses, the detection performance obtained with the automated alignment technique was found to be higher than that obtained with simulated misalignments. Detection performance decreased gradually as the amount of simulated misalignment increased. These results indicate that automatic alignment of breast images is possible and that mass-detection performance appears to improve with the inclusion of asymmetric anatomic information but is not sensitive to slight misalignment.

Recovery of ego-motion using image stabilization

Abstract: A method for computing the 3D camera motion (the ego-motion) in a static scene is introduced, which is based on computing the 2D image motion of a single image region directly from image intensities. The computed image motion of this image region is used to register the images so that the detected image region appears stationary. The resulting displacement field for the entire scene between the registered frames is affected only by the 3D translation of the camera. After canceling the effects of the camera rotation by using such 2D image registration, the 3D camera translation is computed by finding the focus-of-expansion in the translation-only set of registered frames. This step is followed by computing the camera rotation to complete the computation of the ego-motion. The presented method avoids the inherent problems in the computation of optical flow and of feature matching, and does not assume any prior feature detection or feature correspondence. >

Golden template comparison using efficient image registration

Abstract: Golden Template Comparison (GTC) is a method that can be applied to flaw and defect detection in images of 2-dimensional scenes. When a test image is compared to a golden template image, the images must be registered, and then subtracted. The resulting difference image is then analyzed for features that indicate flaws or defects. The registration step is a major determinant of the performance of GTC, and the invention performs the registration step of GTC using a highly efficient and accurate registration method. The registration method of the invention provides substantial registration of all of the features common to the test image and the golden template image, even when one of the images to be registered is flawed, using a registration metric based on minimization of the sum of the absolute value of the differences between the images to be registered, and an efficient optimization technique, such as a version of gradient descent, wherein a local minimum in a registration metric space is found to the nearest pixel using less computational resources than needed to compute the entire registration metric space.

Hierarchical spline-based image registration

Abstract: The problem of image registration subsumes a number of topics in multiframe image analysis, including the computation of optic flow (general pixel-based motion), stereo correspondence, structure from motion, and feature tracking. We present a new registration algorithm based on a spline representation of the displacement field which can be specialized to solve all of the above mentioned problems. In particular, we show how to compute local flow, global (parametric) flow, rigid flow resulting from camera egomotion, and multiframe versions of the above problems. Using a spline-based description of the flow removes the need for overlapping correlation windows, and produces an explicit measure of the correlation between adjacent flow estimates. We demonstrate our algorithm on multiframe image registration and the recovery of 3D projective scene geometry. We also provide results on a number of standard motion sequences. >

Quantitative comparisons of image registration techniques based on high-resolution MRI of the brain.

Abstract: Objective A variety of methods for matching intrasubject MRI-MRI, PET-PET, or MRI-PET image pairs have been proposed. Based on the rigid body transformations needed to align pairs of high-resolution MRI scans and/or simulated PET scans (derived from these MRI scans), we obtained general comparisons of four intrasubject image registration techniques: Talairach coordinates, head and hat, equivalent internal points, and ratio image uniformity. In addition, we obtained a comparison of stereotaxic Z frames with a customized head mold for MRI-MRI image pairs. Materials and methods and results Each technique was quantitatively evaluated using the mean and maximum voxel registration errors for matched voxel pairs within the brain volumes being registered. Conclusion We conclude that fiducial markers such as stereotaxic Z frames that are not rigidly fixed to a patient's skull are inaccurate compared with other registration techniques, Talairach coordinate transformations provide surprisingly good registration, and minimizing the variance of MRI-MRI, PET-PET, or MRI-PET ratio images provides significantly better registration than all other techniques tested. Registration optimization based on measurement of the similarity of spatial distributions of voxel values is superior to techniques that do not use such information.

Accurate frameless registration of MR and CT images of the head: applications in planning surgery and radiation therapy.

Abstract: PURPOSE: To evaluate the feasibility and efficacy of a three-dimensional image registration technique for planning skull base surgery, performing frameless image registration for stereotaxic neurosurgery, and staging nasopharyngeal carcinoma. MATERIALS AND METHODS: Computed tomographic (CT) and magnetic resonance (MR) images from 35 patients were registered by identifying 12-16 homologous landmarks with each modality. Images were displayed as overlaid sections or rendered three-dimensional scenes. The clarity of the combined images from 15 patients undergoing skull base surgery was compared with that of the conventional displays. RESULTS: Images were combined for three applications, with an accuracy of 1-2 mm. For the 15 patients undergoing skull base surgery, the combined images were significantly better at depicting the relationship between bone and lesion than conventional display (P < .01). CONCLUSION: MR and CT images of the head can be accurately registered without using external markers or substant...

Extremal points: definition and application to 3D image registration

Abstract: This paper presents the extraction of feature points, geometrically invariant, which can be used as reliable landmarks for registration and recognition. Mainly we show that it is possible to extract automatically feature lines and points from 3D images, which are robust and invariant with respect to rigid transforms, and enough precise to perform the automatic registration. This extends the possible applications of "the Marching Lines" algorithm, previously described for extremal lines extraction. We introduce here a new kind of feature points: the Extremal Points, belonging to the extremal lines, which allow us to find a point to point correspondences between 3D images of the same subject. We present also experimental results of automatic registration with real data, which demonstrate the remarkable stability of those points. >

Automated 3D nonlinear deformation procedure for determination of gross morphometric variability in human brain

Abstract: We describe an automated method to register MRI volumetric datasets to a digital human brain model. The technique employs3D non-linear warping based on the estimation of local deformation fields using cross-correlation of invariant intensity featuresderived from image data. Results of the non-linear registration on a simple phantom, a complex brain phantom and real MRIdata are presented. Anatomical variability is expressed with respect to the Talairach-like standardized brain-based coordinatesystem of the model. We show that the automated non-linear registration reduces the inter-subject variability of homologouspoints in standardized space by 15% over linear registration methods. A 3D variability map is shown. 1 INTRODUCTION New imaging modalities and techniques, e.g., PET, functional MRI (fMRI), SPECT, magnetoencephalography (MEG), andEEG have made it possible to map functional areas of the human brain with respect to anatomy. Two aspects of this workrequire integration of data from different individuals: 1) The low signal associated with cognitive activation (e.g., a subtlechange in cerebral blood flow (CBF) as measured by PET) requires averaging across subjects to improve statistical significanceofmeasured CBF changes'4"°. 2) Although high resolution imaging techniques such as fMRI now make it possible to measureactivation within a single subject, it will still be necessary to compare results across individuals in order to fully understandthe relationship between functional areas and the underlying gross morphology such as gyral anatomy. For both situations weideally wish to remove all morphological differences between individual brains before considering the distribution of functionalinformation superimposed on the anatomical substrate. This requires a method for deforming one brain to match another atall points, and has typically been accomplished by mapping the volumetric data into a standardized brain-based coordinatesystem24. Until recently, most centers have used linear transformations only13'15'19'25. However, previous work24'23 has shownthat even after linear mapping, there is variability of up to 1.5 cm in the position of cortical structures, which may representa significant source of error when mapping activation foci. We have shown9, that on average for points throughout the brain(cortical and sub-cortical), there is a 6-7mm anatomical variability in 3D position not accounted for by linear registration.The objective of this paper is to present an automated method of establishing the non-linear morphometric variability ina population of normal brains with 3D MRI. Non-linear warping based on homologous landmark matching2'9 has not beenpractical for routine use as a deformation/warping model because of the subjectivity involved in selecting the precise locationand number of points that define the non-linear deformation. This has lead our group and others (e.g.,1"6'20) to consider fullyautomated, objective non-linear mapping techniques. Our method uses non-linear 3D warping of one data set to register itwith another, based on the estimation of local deformations derived from local neighbourhood correlation of invariant featurescalculated from image data3'4.To properly assess non-linear variability it is first essential to have a well-defined 3D coordinate space where the linearcomponent of the anatomical variation is removed by application of an affine transformation. Without a priori knowledgeof anatomical variability, the best minimum variance frame cannot be defined since it is wholly dependent on the former.Therefore, we have selected a brain-based coordinate system very similar to that proposed by Talairach24. Our implementationuses a single global affine transformation whereas Talairach employs 12 piece-wise linear transformations (as implemented in

Building 3-D models from unregistered range images

Abstract: The authors describe an approach to building a three-dimensional model from a set of range images. The authors' goal is to build models of free-form surfaces obtained from arbitrary viewing directions, with no initial estimate of the relative viewing directions. The approach is based on building discrete meshes representing the surfaces observed in each of the range images, to map each of the meshes to a spherical image, and to compute the transformations between the views by matching the spherical images. The meshes are built using an iterative fitting algorithm previously developed; the spherical images are built by matching the nodes of the surface meshes to the nodes of a reference mesh on the unit sphere and by storing a measure of curvature at every node. The authors describe the algorithms used for building such models from range images and for matching them. The authors give results obtained using range images of complex objects. >

Registration of multiple range views for automatic 3-D model building

Abstract: Building integrated models of existing 3-D objects is a key requirement for both reverse engineering and object recognition systems. An automatic 3-D model builder goes through three main steps: i) surface sampling from many views, ii) registration of the sampled views, and iii) integration of the registered views. The accuracy obtained depends on the acquisition and registration errors. The latter is critical since a misalignment of the range views causes their noise distributions to be centered around different means, which makes it difficult to reduce the effect of the acquisition error by simple averaging. In this paper, we propose a general algorithm that reduces significantly the level of the registration errors between all pairs in a set of range views. This algorithm refines initial estimates of the transformation matrices obtained from the calibrated acquisition setup. It considers the network of views as a whole and minimizes the registration errors of all views simultaneously. This leads to a well-balanced network of views in which the registration errors are equally distributed. Experimental results show an improvement of both the calibrated registrations and integrated models. >

Point correlation: a reduced-cost template matching technique

Abstract: The registration of a scene w.r.t. a reference template is usually performed by computing the extremum of the correlation surface. Unfortunately, conventional correlation-type algorithms are computationally expensive. We propose a new reduced-cost correlation technique (called "point correlation") where matching is not performed with the entire template but with a precomputed set of points of this template. We introduce a method to iteratively select a set of points which is optimal in a heuristic way. >

Three-dimensional image registration method for spiral CT angiography

Abstract: A three-dimensional image registration method also uses an algorithm and applies it to the solution of such problems in 3D CT DSA. The method can deal with incomplete or partial volumetric data in registration and correct patient global motion prior to subtraction even when it is coupled with local unconscious/nonrigid movements. Experimental results demonstrate the effectiveness of this algorithm on several clinical spiral CT data.

Reconstruction of a high-resolution image from multiple-degraded misregistered low-resolution images

Abstract: In applications that demand highly detailed images, it is often not feasible or sometimes possible to acquire images of such high resolution by just using hardware (high precision optics and charge coupled devices). Instead, image processing approaches can be used to construct a high resolution image from multiple, degraded, low resolution images. It is assumed that the low resolution images have been subsampled (thus introducing aliasing) and displaced by sub-pixel shifts with respect to a reference frame. Therefore, the problem can be divided into three sub-problems: registration (estimating the shifts), restoration, and interpolation. None of the methods which appeared in the literature solve the registration and restoration sub-problems simultaneously. This is sub-optimal, since the registration and restoration steps are inter-dependent. Based on previous restoration and identification work using the Expectation-Maximization algorithm, the proposed approach estimates the sub-pixel shifts and conditional mean (restored images) simultaneously. In addition, the registration and restoration sub-problems are cast in a multi-channel framework to take advantage of the cross- channel information. Experimental results show the validity of this method.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Parallel registration of multisensor remotely sensed imagery using wavelet coefficients

Abstract: Due to the increasing amount and diversity of remotely sensed data, image registration is becoming one of the most important issues in remote sensing. In the near future, remote sensing systems will provide large amounts of data representing multiple- time or simultaneous observations of the same features by different sensors. The combination of data from coarse-resolution satellite sensors designed for large-area survey and from finer- resolution sensors for more detailed studies will allow better analysis of each type of data as well as validation of global low-resolution data analysis by the use of local high-resolution data analysis. This integration of information from multiple sources starts with the registration of the data. The most common approach to image registration is to choose, in both input image and reference image, some well-defined ground control points (GCPs), and then to compute the parameters of a deformation model. The main difficulty lies in the choice of the GCPs. In our work, a parallel implementation of decomposition and reconstruction by wavelet transforms has been developed on a single-instruction multiple-data (SIMD) massively parallel computer, the MasPar MP-1. Utilizing this framework, we show how maxima of wavelet coefficients, which can be used for finding ground control points of similar resolution remotely sensed data, can also form the basis of the registration of very different resolution data, such as data from the NOAA Advanced Very High Resolution Radiometer (AVHRR) and from the Landsat/Thematic Mapper (TM).

Efficient image registration

Abstract: A method and apparatus for efficient registration of a pair of digitized images is provided that obtains a registration metric value based upon a Sum of Absolute Differences registration metric computation for each of a plurality of neighboring-pixel relative displacements, and, for example, iteratively selects a new initial relative displacement from among the plurality of neighboring-pixel relative displacements such that each succeeding new initial relative displacement is associated with a smaller registration metric value, until an initial relative displacement that is associated with a minimum registration metric value is reached. In general, the relative displacement that is associated with the minimum registration metric value is located using a two-dimensional numerical optimization analysis. The invention is especially useful for flaw and defect analysis, such as Golden Template Analysis, third optical inspection, as well as for pair-wise comparison of die images on a semiconductor wafer.

Reconstruction of a high resolution image from registration and restoration of low resolution images

Abstract: In applications that demand highly detailed images, it is often not feasible nor sometimes possible to acquire images of such high resolution by just using hardware (high precision optics and charge coupled devices (CCDs)). Instead, image processing methods may be used to construct a high resolution image from multiple, degraded, low resolution images. It is assumed that the low resolution images have been sub-sampled and displaced by sub-pixel shifts. Therefore, the problem can be divided into three sub-problems: registration (estimating the shifts), restoration, and interpolation. This paper focuses on solving the first two sub-problems simultaneously, using the expectation-maximization (EM) algorithm. Experimental results are presented that demonstrate the effectiveness of this approach. >

Brain image registration based on curve mapping

Abstract: A new two-stage approach for brain image registration is proposed. In the first stage, an active contour algorithm is used to establish a length-preserving, one-to-one mapping between the cortical and the ventricular boundaries in the two images to be registered. This mapping is used in the second step by a two-dimensional transformation which is based on an elastic body deformation. This method was tested by registering magnetic resonance images to both photographic pathology images and atlas images. >

A general technique for interstudy registration of multifunction and multimodality images

Abstract: A technique that can register anatomic/structural brain images (e.g., MRI) with various functional images (e.g., PET-FDG and PET-FDOPA) of the same subject has been developed. The procedure of this technique includes the following steps: (1) segmentation of MRI brain images into gray matter (GM), white matter (WM), cerebral spinal fluid (CSF), and, muscle (MS) components, (2) assignment of appropriate radio-tracer concentrations to various components depending on the kind of functional image that is being registered, (3) generation of simulated functional images to have a spatial resolution that is comparable to that of the measured ones, (4) alignment of the measured functional images to the simulated ones that are based on MRI images. A self-organization clustering method is used to segment the MRI images. The image alignment is based on the criterion of least squares of the pixel-by-pixel differences between the two sets of images that are being matched and on the Powell's algorithm for minimization. The technique was applied successfully for registering the MRI, PET-FDG, and PET-FDOPA images. This technique offers a general solution to the registration of structural images to functional images and to the registration of different functional images of markedly different distributions. >

A semiautomatic method for registration of portal images

Abstract: A method to register pairs of portal images quickly and accurately has been developed. The approach uses a cross-correlation operator to find the optimal match between corresponding anatomic regions that have been selected by a user on pairs of portal images. The cross-correlation operator determines the translation that best registers each pair of anatomic features independently, and then the images are translated, rotated, and scaled so that the least squares difference between the coordinates of all of the paired regions is minimized. Tests using simulated images have shown that the accuracy of the algorithm is dependent on (i) the size and shape of the structures within the paired regions; (ii) the subject contrast of the anatomic features being matched; (iii) the rotational difference between the anatomic features; and, (iv) the noise in the image. Tests using contrast-detail phantom images have shown that the semiautomated registration algorithm is only slightly less accurate than human observers, but is considerably faster. The semiautomated algorithm shows great promise as a method of quickly and accurately estimating discrepancies in patient positioning when used in conjunction with an on-line portal imaging system.

Matching structural images of the human brain using statistical and geometrical image features

Abstract: The efficacy of using intensity edges, curvature of iso-intensity contours, and tissue classified data for image matching are examined. The image matching problem is formulated in such a way that the different features are handled uniformly, allowing the same code to be used in each instance. The results using both simulated and real brain images indicate that each feature affected and improvement in the correspondence after matching with it.