• Journal of the Semiconductor & Display Technology
• /
• v.18 no.1
• /
• pp.102-104
• /
• 2019

Registering different kinds of clinical images widely used in diagnostic and surgery planning. However, cause of tumor growth or effected by gravity, human tissue has plenty of non-rigid deformation with clinically. Non-rigid registration allows the mapping of straight lines to curves. Therefore, such local deformation makes registration more complicated. In this work, we mainly introduce intra-subject, inter-modality registration. This paper mainly studies the nonlinear registration method of 2D medical image registration. The general medical image registration algorithm requires manual intervention, and cost long registration time. In our work to reduce the registration time in rough registration step, the barycenter and the direction of main axis of the image is calculated, which reduces the calculation amount compared with the method of using mutual information.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.1
• /
• pp.180-194
• /
• 2021

For small image registration, feature-based approaches are likely to fail as feature detectors cannot detect enough feature points from low-resolution images. The classic FFT approach's prediction accuracy is high, but the registration time can be relatively long, about several seconds to register one image pair. To achieve real-time and high-precision rigid registration for small images, we apply deep neural networks for supervised rigid transformation prediction, which directly predicts the transformation parameters. We train deep registration models with rigidly transformed CIFAR-10 images and STL-10 images, and evaluate the generalization ability of deep registration models with transformed CIFAR-10 images, STL-10 images, and randomly generated images. Experimental results show that the deep registration models we propose can achieve comparable accuracy to the classic FFT approach for small CIFAR-10 images (32×32) and our LSTM registration model takes less than 1ms to register one pair of images. For moderate size STL-10 images (96×96), FFT significantly outperforms deep registration models in terms of accuracy but is also considerably slower. Our results suggest that deep registration models have competitive advantages over conventional approaches, at least for small images.

• Journal of Biomedical Engineering Research
• /
• v.24 no.4
• /
• pp.259-266
• /
• 2003

Non-rigid registration between different modality images with shape deformation can be used to diagnosis and study for inter-patient image registration, longitudinal intra-patient registration, and registration between a patient image and an atlas image. This paper proposes a hierarchical registration method using bodily tissue based segmentation for registration between color images and CT images of the Visible Human leg areas. The cross-sectional color images and the axial CT images are segmented into three distinctive bodily tissue regions, respectively: fat, muscle, and bone. Each region is separately registered hierarchically. Bounding boxes containing bodily tissue regions in different modalities are initially registered. Then, boundaries of the regions are globally registered within range of searching space. Local boundary segments of the regions are further registered for non-rigid registration of the sampled boundary points. Non-rigid registration parameters for the un-sampled points are interpolated linearly. Such hierarchical approach enables the method to register images efficiently. Moreover, registration of visibly distinct bodily tissue regions provides accurate and robust result in region boundaries and inside the regions.

• Journal of Korea Multimedia Society
• /
• v.16 no.6
• /
• pp.700-707
• /
• 2013

In this paper, we propose a non-rigid registration method of lung parenchyma in temporal chest CT scans using region binarization modeling and locally deformable model. To cope with intensity differences between CT scans, we segment the lung vessel and parenchyma in each scan and perform binarization modeling. Then, we match them without referring any intensity information. We globally align two lung surfaces. Then, locally deformable transformation model is developed for the subsequent non-rigid registration. Subtracted quantification results after non-rigid registration are visualized by pre-defined color map. Experimental results showed that proposed registration method correctly aligned lung parenchyma in the full inspiration and expiration CT images for ten patients. Our non-rigid lung registration method may be useful for the assessment of various lung diseases by providing intuitive color-coded information of quantification results about lung parenchyma.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.15 no.5
• /
• pp.64-74
• /
• 2019

Although a non-rigid registration has high demands in clinical practice, it has a high computational complexity and it is very difficult for ensuring the accuracy and robustness of registration. This study proposes a method of applying a non-rigid registration to 3D magnetic resonance images of brain in an unsupervised learning environment by using a deep-learning network. A feature vector between two images is produced through the network by receiving both images from two different patients as inputs and it transforms the target image to match the source image by creating a displacement vector field. The network is designed based on a U-Net shape so that feature vectors that consider all global and local differences between two images can be constructed when performing the registration. As a regularization term is added to a loss function, a transformation result similar to that of a real brain movement can be obtained after the application of trilinear interpolation. This method enables a non-rigid registration with a single-pass deformation by only receiving two arbitrary images as inputs through an unsupervised learning. Therefore, it can perform faster than other non-learning-based registration methods that require iterative optimization processes. Our experiment was performed with 3D magnetic resonance images of 50 human brains, and the measurement result of the dice similarity coefficient confirmed an approximately 16% similarity improvement by using our method after the registration. It also showed a similar performance compared with the non-learning-based method, with about 10,000 times speed increase. The proposed method can be used for non-rigid registration of various kinds of medical image data.

• Korean Journal of Remote Sensing
• /
• v.37 no.3
• /
• pp.475-491
• /
• 2021

With the recent increase in available high-resolution (< ~1 m) satellite SAR images, the demand for precise registration of SAR images is increasing in various fields including change detection. The registration between high-resolution SAR images acquired in different look angle is difficult due to speckle noise and geometric distortion caused by the characteristics of SAR images. In this study, registration is performed in two stages, coarse and fine, using the x-band SAR data imaged at staring spotlight mode of TerraSAR-X. For the coarse registration, a method combining the adaptive sampling method and SAR-SIFT (Scale Invariant Feature Transform) is applied, and three rigid methods (NCC: Normalized Cross Correlation, Phase Congruency-NCC, MI: Mutual Information) and one non-rigid (Gefolki: Geoscience extended Flow Optical Flow Lucas-Kanade Iterative), for the fine registration stage, was performed for performance comparison. The results were compared by using RMSE (Root Mean Square Error) and FSIM (Feature Similarity) index, and all rigid models showed poor results in all image combinations. It is confirmed that the rigid models have a large registration error in the rugged terrain area. As a result of applying the Gefolki algorithm, it was confirmed that the RMSE of Gefolki showed the best result as a 1~3 pixels, and the FSIM index also obtained a higher value than 0.02~0.03 compared to other rigid methods. It was confirmed that the mis-registration due to terrain effect could be sufficiently reduced by the Gefolki algorithm.

• Journal of Korea Multimedia Society
• /
• v.18 no.10
• /
• pp.1157-1163
• /
• 2015

In this paper, we propose a registration method of 3d facial scan data and CBCT data using voxelization and distance map. First, two data sets are initially aligned by exploiting the voxelization of 3D facial scan data and the information of the center of mass. Second, a skin surface is extracted from 3D CBCT data by segmenting air and skin regions. Third, the positional and rotational differences between two images are accurately aligned by performing the rigid registration for the distance minimization of two skin surfaces. Experimental results showed that proposed registration method correctly aligned 3D facial scan data and CBCT data for ten patients. Our registration method might give useful clinical information for the oral surgery planning and the diagnosis of the treatment effects after an oral surgery.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.10C
• /
• pp.1402-1413
• /
• 2004

In non-rigid image registration, the Jacobian determinant of the estimated deformation should be positive everywhere since physical deformations are always invertible. We propose a constrained optimization technique at ensures the positiveness of Jacobian determinant for cubic B-spline based deformation. We derived sufficient conditions for positive Jacobian determinant by bounding the differences of consecutive coefficients. The parameter set that satisfies the conditions is convex; it is the intersection of simple half spaces. We solve the optimization problem using a gradient projection method with Dykstra's cyclic projection algorithm. Analytical results, simulations and experimental results with inhale/exhale CT images with comparison to other methods are presented.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.52 no.8
• /
• pp.500-505
• /
• 2003

In this paper, the method to register multiple sets of skull CT images to absolute coordinate system is proposed. Contrary to correspondence paired mapping of previous techniques, four anatomical landmark points, three coplanar points and one non-coplanar point, compose three principal axes simple and unique for efficient registration by means of rigid body transformation. Throughout the numerical simulation with added random noises, the error performances in terms of different rotation and rounding-off of landmark points, and incorrect localization of anatomical landmark and target points are quantitatively analyzed to generalize the proposed technique. Experiments using real skull CT images demonstrate the feasibility for an efficient use in clinical practice.

• Journal of Korea Multimedia Society
• /
• v.16 no.12
• /
• pp.1454-1464
• /
• 2013

In this paper, we propose a rapid rigid registration method for the fusion visualization of intra-operative 2D XA and pre-operative 3D CTA images. In this paper, we propose a global movement estimation based on a trilateration for the fast and robust initial registration. In addition, the principal axis of each image is generated and aligned, and the bounding box of the vascular shape is compared for more accurate initial registration. For the fine registration, two images are registered where the distance between two vascular structures is minimized by selective distance measure. In the experiment, we evaluate a speed, accuracy and robustness using five patients' data by comparing the previous registration method. Our proposed method shows that two volumes can be registered at optimal location rapidly, and robustly comparing with the previous method.