On-line range images registration with GPGPU

This paper concerns implementation of algorithms in the two important aspects of modern 3D data processing: data registration and segmentation. Solution proposed for the first topic is based on the 3D space decomposition, while the latter on image processing and local neighbourhood search. Data processing is implemented by using NVIDIA compute unified device architecture (NIVIDIA CUDA) parallel computation. The result of the segmentation is a coloured map where different colours correspond to different objects, such as walls, floor and stairs. The research is related to the problem of collecting 3D data with a RGB-D camera mounted on a rotated head, to be used in mobile robot applications. Performance of the data registration algorithm is aimed for on-line processing. The iterative closest point (ICP) approach is chosen as a registration method. Computations are based on the parallel fast nearest neighbour search. This procedure decomposes 3D space into cubic buckets and, therefore, the time of the matching is deterministic. First technique of the data segmentation uses accele-rometers integrated with a RGB-D sensor to obtain rotation compensation and image processing method for defining pre-requisites of the known categories. The second technique uses the adapted nearest neighbour search procedure for obtaining normal vectors for each range point.     

Effect of nonrigid registration algorithms on deformation-based morphometry: a comparative study with control and Williams syndrome subjects

Deformation-based morphometry (DBM) is a widely used method for characterizing anatomical differences across groups. DBM is based on the analysis of the deformation fields generated by nonrigid registration algorithms, which warp the individual volumes to a DBM atlas. Although several studies have compared nonrigid registration algorithms for segmentation tasks, few studies have compared the effect of the registration algorithms on group differences that may be uncovered through DBM. In this study, we compared group atlas creation and DBM results obtained with five well-established nonrigid registration algorithms using 13 subjects with Williams syndrome and 13 normal control subjects. The five nonrigid registration algorithms include the following: (1) the adaptive bases algorithm, (2) the image registration toolkit, (3) The FSL nonlinear image registration tool, (4) the automatic registration tool, and (5) the normalization algorithm available in Statistical Parametric Mapping (SPM8). Results indicate that the choice of algorithm has little effect on the creation of group atlases. However, regions of differences between groups detected with DBM vary from algorithm to algorithm both qualitatively and quantitatively. Some regions are detected by several algorithms, but their extent varies. Others are detected only by a subset of the algorithms. Based on these results, we recommend using more than one algorithm when performing DBM studies.     

Real-time extended-field-of-view ultrasound based on a standard PC

Extended field-of-view (EFOV) can acquire a full field of vision, which can help doctors to make more objective and accurate diagnosis. Current EFOV techniques suffer from the low computation speed due to the large amount of ultrasound data to be processed. This paper describes an efficient technique to register 2D multiframe ultrasound images and produce EFOV images with significantly reduced computation time based on a standard PC. For registration of any two adjacent images, we propose to select less image blocks which are regarded as the most valid blocks based on the importance of image content. In registration of a sequence of images, with an assumption that the moving direction and speed of the probe are nearly identical during the data collection, we estimate the moving speed of the probe at the beginning of data collection and ignore redundant image data by processing a smaller number of frames according to a frame interval. The experimental results show that the computation speed of our method is increased by 7–80 times in comparison with two traditional methods, and can accurately produce EFOV images in real-time.     

Separable convolution template (SCT) background prediction accelerated by CUDA for infrared small target detection

This paper presents a novel background prediction method for infrared small target detection (ISTD). Using a separable convolution template (SCT) to accelerate the traditional background prediction by graphic processing unit (GPU), the new method provides a significant improvement in the prediction speed, which enables the prediction process in real time. And experimental results show its high efficiency and practical application over previous work. The mathematical approach proposed here could be extended to accelerate the applications referred to image convolutions not only to the infrared field.     

Combining Sparse and Dense Features to Improve Multi-Modal Registration for Brain DTI Images

A new solution to overcome the constraints of multimodality medical intra-subject image registration is proposed, using the mutual information (MI) of image histogram-oriented gradients as a new matching criterion. We present a rigid, multi-modal image registration algorithm based on linear transformation and oriented gradients for the alignment of T2-weighted (T2w) images (as a fixed reference) and diffusion tensor imaging (DTI) (b-values of 500 and 1250 s/mm²) as floating images of three patients to compensate for the motion during the acquisition process. Diffusion MRI is very sensitive to motion, especially when the intensity and duration of the gradient pulses (characterized by the b-value) increases. The proposed method relies on the whole brain surface and addresses the variability of anatomical features into an image stack. The sparse features refer to corners detected using the Harris corner detector operator, while dense features use all image pixels through the image histogram of oriented gradients (HOG) as a measure of the degree of statistical dependence between a pair of registered images. HOG as a dense feature is focused on the structure and extracts the oriented gradient image in the x and y directions. MI is used as an objective function for the optimization process. The entropy functions and joint entropy function are determined using the HOGs data. To determine the best image transformation, the fiducial registration error (FRE) measure is used. We compare the results against the MI-based intensities results computed using a statistical intensity relationship between corresponding pixels in source and target images. Our approach, which is devoted to the whole brain, shows improved registration accuracy, robustness, and computational cost compared with the registration algorithms, which use anatomical features or regions of interest areas with specific neuroanatomy. Despite the supplementary HOG computation task, the computation time is comparable for MI-based intensities and MI-based HOG methods.     

Simulation of Properties of Images with Atomic Resolution in a Scanning Probe Microscope

A method for simulation of images in a scanning probe microscope (SPM) using simultaneous wavelet transform and median filtering is proposed. The wavelet transform with the fourth-order Daubechies kernel is used. Such a transform makes it possible to select details of different scales in the SPM image and, hence, study fractal properties of surfaces. Simulation is used to show that ultrahigh (atomic) resolution is possible in SPM provided that the size of the contact region in the probe–sample system is significantly greater than atomic size and the lattice atoms are randomly distributed. Contrast inversion in the SPM images in the multiscan mode is interpreted.     

How does distortion correction correlate with anisotropic indices? A diffusion tensor imaging study

PURPOSE: The purpose of this study was to determine a suitable registration algorithm for diffusion tensor imaging (DTI) using conventional preprocessing tools [statistical parametric mapping (SPM) and automated image registration (AIR)] and to investigate how anisotropic indices for clinical assessments are affected by these distortion corrections. MATERIALS AND METHODS: Brain DTI data from 15 normal healthy volunteers were used to evaluate four spatial registration schemes within subjects to correct image distortions: noncorrection, SPM-based affine registration, AIR-based affine registration and AIR-based nonlinear polynomial warping. The performance of each distortion correction was assessed using: (a) quantitative parameters: tensor-fitting error (Ef), mean dispersion index (MDI), mean fractional anisotropy (MFA) and mean variance (MV) within 11 regions of interest (ROI) defined from homogeneous fiber bundles; and (b) fiber tractography through the uncinate fasciculus and the corpus callosum. Fractional anisotropy (FA) and mean diffusivity (MD) were calculated to demonstrate the effects of distortion correction. Repeated-measures analysis of variance was used to investigate differences among the four registration paradigms. RESULTS: AIR-based nonlinear registration showed the best performance for reducing image distortions with respect to smaller Ef (P<.02), MDI (P<.01) and MV (P<.01) with larger MFA (P<.01). FA was decreased to correct distortions (P<.0001) whether the applied registration was linear or nonlinear and was lowest after nonlinear correction (P<.001). No significant differences were found in MD. CONCLUSION: In conventional DTI processing, anisotropic indices of FA can be misestimated by noncorrection or inappropriate distortion correction, which leads to an erroneous increase in FA. AIR-based nonlinear distortion correction would be required for a more accurate measurement of this diffusion parameter.     

Real-time automatic registration in optical surgical navigation

An image-guided surgical navigation system requires the improvement of the patient-to-image registration time to enhance the convenience of the registration procedure. A critical step in achieving this aim is performing a fully automatic patient-to-image registration. This study reports on a design of custom fiducial markers and the performance of a real-time automatic patient-to-image registration method using these markers on the basis of an optical tracking system for rigid anatomy. The custom fiducial markers are designed to be automatically localized in both patient and image spaces. An automatic localization method is performed by registering a point cloud sampled from the three dimensional (3D) pedestal model surface of a fiducial marker to each pedestal of fiducial markers searched in image space. A head phantom is constructed to estimate the performance of the real-time automatic registration method under four fiducial configurations. The head phantom experimental results demonstrate that the real-time automatic registration method is more convenient, rapid, and accurate than the manual method. The time required for each registration is approximately 0.1 s. The automatic localization method precisely localizes the fiducial markers in image space. The averaged target registration error for the four configurations is approximately 0.7 mm. The automatic registration performance is independent of the positions relative to the tracking system and the movement of the patient during the operation.     

A least squares quantization table method for direct reconstruction of MR images with non-Cartesian trajectory

The direct Fourier transform method is a straightforward solution with high accuracy for reconstructing magnetic resonance (MR) images from nonuniformly sampled k-space data, given that the optimal density compensation function is selected and the underlying magnetic field is sufficiently uniform. The computation however is very time-consuming, making it impractical especially for large-size images. In this paper, the least squares quantization table (LSQT) method is proposed to accelerate the direct Fourier transform computation, similar to the recently proposed methods such as using look-up table (LUT) or equal-phase-line (EPL). With LSQT, all the image pixels are first classified into several groups where the Lloyd-Max quantization scheme is used to ensure the minimal classification error. The representative value of each group is stored in a small-size LSQT in advance to reduce the computational load. The pixels in the same group receive the same contribution, which is calculated only once for each group instead of for each pixel, resulting in the reduction of computation because the number of groups is far smaller than the number of pixels. Finally, each image pixel is mapped into the nearest group and its representative value is used to reconstruct the image. The experimental results show that the LSQT method requires far smaller memory size than the LUT method and fewer multiplication operations than the LUT and EPL methods. Moreover, the LSQT method can perform large-size reconstructions that achieve comparable or higher accuracy as compared to the EPL and gridding methods when the appropriate parameters are given. The inherent parallel structure also makes the LSQT method easily adaptable to a multiprocessor system.     

Multispectral registration method based on stellar trajectory fitting

Image registration techniques for detecting moving targets in space require high temporal resolution and multispectral images to improve the target detection probability and reduce the false positive rate. At present, image registration accuracy is affected by the effective number of common multispectral registration control points as well as its stability. Image registration based on on-orbit stellar trajectory measurement can be used to perform on-orbit modifications of registration deviation caused by thermal distortions during launch. This study proposes a new image registration method based on the on-orbit detection of stars by using the stellar trajectory on a camera’s focal plane. A generated 12?×?12 data template and the Lagrange interpolation method are used in the registration model. Multispectral image registration based on stellar trajectory fitting can achieve high-precision image registration among different spectra.