Image Registration by Template Matching Using Normalized Cross-Correlation

For many image processing applications (such as feature tracking, object recognition, stereo matching and remote sensing), the technique known as Template Matching (TM) plays an important role for the localization and recognition of objects or patterns within a digital image. A TM approach seeks to find a position within a source image which yields to the best possible resemblance between a given sub-image (typically referred as image template) and a corresponding region of such source image. TM involves two critical aspects: similarity measurement and search strategy. In this sense, the simplest available TM method involves an exhaustive computation of the Normalized Cross-Correlation (NCC) value (similarity measurement) over all pixel locations of the source image (search strategy). Unfortunately, this approach is strongly restricted due to the high computational cost implied in the evaluation of the NCC coefficient. Recently, several TM methods based on evolutionary approaches have been proposed as an alternative to reduce the number of search locations in the TM process. However, the lack of balance between exploration and exploitation related to the operators employed by many of such approaches makes TM to suffer from several critical flaws, such as premature convergence. In the proposed approach, the swarm optimization method known as Locust Search (LS) is applied to solve the problem of template matching. The unique evolutionary operators employed by LS method’s search strategy allows to explicitly avoid the concentration of search agents toward the best-known solutions, which in turn allows a better exploration of the valid image’s search region. Experimental results show that, in comparison to other similar methods, the proposed approach achieves the best balance between estimation accuracy and computational cost.

In many image processing applications, the Template Matching (TM) technique plays an important role in recognizing and locating patterns or objects within a digital image. The main task of the TM is to seeks and find a position within an original image that resembles a predetermined sub-image (template) and a corresponding region of the original image as much as possible. TM involves two main aspects: similarity measurement and search strategy. The simplest method applied to TM involves a comprehensive calculation of the value of the Normalized Cross-Correlation (NCC) over all the pixel locations of the source image (search strategy). Unfortunately, the high computational cost that implies the evaluation of the NCC coefficient makes this approach restricted. We propose several TM methods based on evolutionary approaches as an alternative to reduce the number of search locations in the TM process. We have conducted a comparison of several evolutionary methods, to obtain which of these is the optimum to perform the TM task. Experimental results of this comparison show us which methods achieve the best balance between estimation and computational cost.

In digital image processing, template matching is a process to determine the location of sub image inside an image. The sub image, which is called template, usually has similarity with a part of the image. The template can be in different size, color or form. Template matching is famously used in image registration and object recognition. In this paper, we focus on the performance of the Sum of Squared Differences (SSD) and Normalized Cross Correlation (NCC)as the techniques that used in image registration for matching the template with an image. This experiment is aiming to compare the ability of both techniques in term of quality of output image as well as the time taken in execution process. Furthermore, it also to test the effect of template image to output image when there is noise and rotation. This paper provides an explanation about the concept of similarity measure techniques and how the algorithms are implementing in image registration process. Finally, the performance of these methods is tested by making comparison based on the value of correlation coefficient that produced from different image templates.

Purpose: To develop and evaluate a template matching algorithm for direct detection of a lung tumor trajectory in the raw projections acquired from a cone beam CT. Method and Materials: A planning respiration‐correlated CT (pRCCT) was used to construct digitally‐reconstructed radiographic (DRR) templates for registration with raw projection data of a free‐breathing cone beam CT (FB‐CBCT). The pRCCT and FB‐CBCT were first registered to obtain the approximate mean 3D position of the tumor. This position was applied as a shift to the planning isocenter in the pRCCT. Next, 72 DRRs were generated at 5 degree angular increments around this shifted position, for each phase in the pRCCT. The appropriate phase DRR template for registration was selected using the measured phase of the FB‐CBCT projection. This phase was measured by tracking the diaphragm apex position in the projections. Two template matching algorithms were compared: normalized cross correlation and a block normalized cross correlation. A dynamic phantom was induced to move reproducibly 2cm during planning and 3cm during online CBCT imaging. The phantom was imaged at isocenter and with a 1cm translational shift from isocenter. Results: The tumor excursion measured with the normalized cross correlation method was 28.7 mm, and 28.5 mm with the block normalized cross correlation. The error in detecting the daily standard deviation of tumor position was 0.1mm for both methods. The mean absolute difference in detected tumor position was 1.07 mm, with a 95% confidence value of 2.43 mm. Block normalized cross correlation reduced the maximum error in tumor position detection slightly, in relation to normalized cross correlation. Conclusion: Direct detection of the tumor position from cone beam CT projections is feasible with template matching, provided that appropriate and accurate templates are used. Conflict of Interest: Supported in part by NIH R01 CA116249 and Elekta, Inc.

Dental radiographs are used to diagnose and treat dental diseases. For accurate diagnosis, geometrical registration of radiographs is required. In order to compare two dental radiographs prepared from the same part of teeth in different time intervals. It is necessary to register the locality of images, so that by removing deviations between the two images and finding a mapping function, local transitions, rotational and scale differences and even visual differences are corrected. In this paper, an X-ray dental image registration algorithm based on template matching and similarity measure is proposed. The proposed registration process is carried out in three stages. The first stage includes the automatic selection of a set of control points of the sensed image based on its features. In the second stage, displacement vectors of these control points are determined by comparing a template and normalized cross correlation (NCC) as a similarity measure. Full displacement vector field is then performed using these displacement vectors and the affine transformation function, and the final correction is carried out by changing the geometric shape of the reference image. The obtained results of the proposed algorithm show effectiveness of this algorithm in dental radiograph image registration.

Template matching has many applications in signal processing, image processing, pattern recognition, and video compression. This paper proposes a fast coarse-to-fine template matching algorithm for finding the exact best match, i.e., the match that may be found by a full search. This is obtained by pruning the number of candidates in the full search using the results of a coarse search. Experimental results show that speed ups of a couple of orders of magnitude can easily be achieved using this method for typical low-noise cases of two-dimensional (2-D) template matching.

Many image registration algorithms rely on the use of homologous control points on the two input image sets to be registered. In reality, the interactive identification of the control points on both images is tedious, difficult, and often a source of error. We propose a two-step algorithm to automatically identify homologous regions that are used as a priori information during the image registration procedure. First, a number of small control volumes having distinct anatomical features are identified on the model image in a somewhat arbitrary fashion. Instead of attempting to find their correspondences in the reference image through user interaction, in the proposed method, each of the control regions is mapped to the corresponding part of the reference image by using an automated image registration algorithm. A normalized cross-correlation (NCC) function or mutual information was used as the auto-mapping metric and a limited memory Broyden-Fletcher-Goldfarb-Shanno algorithm (L-BFGS) was employed to optimize the function to find the optimal mapping. For rigid registration, the transformation parameters of the system are obtained by averaging that derived from the individual control volumes. In our deformable calculation, the mapped control volumes are treated as the nodes or control points with known positions on the two images. If the number of control volumes is not enough to cover the whole image to be registered, additional nodes are placed on the model image and then located on the reference image in a manner similar to the conventional BSpline deformable calculation. For deformable registration, the established correspondence by the auto-mapped control volumes provides valuable guidance for the registration calculation and greatly reduces the dimensionality of the problem. The performance of the two-step registrations was applied to three rigid registration cases (two PET-CT registrations and a brain MRI-CT registration) and one deformable registration of inhale and exhale phases of a lung 4D CT. Algorithm convergence was confirmed by starting the registration calculations from a large number of initial transformation parameters. An accuracy of approximately 2 mm was achieved for both deformable and rigid registration. The proposed image registration method greatly reduces the complexity involved in the determination of homologous control points and allows us to minimize the subjectivity and uncertainty associated with the current manual interactive approach. Patient studies have indicated that the two-step registration technique is fast, reliable, and provides a valuable tool to facilitate both rigid and nonrigid image registrations.

Template matching is a fundamental problem in pattern recognition, which has wide applications, especially in industrial inspection. In this paper, we propose a 1-D template matching algorithm which is an alternative for 2-D full search block matching algorithms. Our approach consists of three steps. In the first step the images are converted from 2-D into 1-D by summing up the intensity values of the image in two directions horizontal and vertical. In the second step, the template matching is performed among 1-D vectors using the similarity function sum of square difference. Finally, the decision will be taken based on the value of similarity function. Transformation template image and sub-images in the source image from 2-D grey level information into 1-D information vector reduce the dimensionality of the data and accelerate the computations. Experimental results show that the computational time of the proposed approach is faster and performance is better than three basic template matching methods. Moreover, our approach is robust to detect the target object with changes of illumination in the template also when the Gaussian noise added to the source image.

As one of the critical steps in brain imaging analysis and processing, brain image registration plays a significant role. In this paper, we proposed a technique of human brain image registration based on tissue morphology in vivo to address the problems of previous image registration. First, different feature points were extracted and combined, including those at the boundary of different brain tissues and those of the maximum or minimum from the original image. Second, feature points were screened through eliminating their wrong matching pairs between moving image and reference image. Finally, the remaining matching pairs of feature points were used to generate the model parameters of spatial transformation, with which the brain image registration can be finished by combining interpolation techniques. Results showed that compared with the Surf, Demons, and Sift algorithms, the proposed method can perform better not only for four quantitative indicators (mean square differences, normalized cross correlation, normalized mutual information and mutual information) but also in spatial location, size, appearance contour, and registration details. The findings may suggest that the proposed method will be of great value for brain image reconstruction, fusion, and statistical comparison analysis.

A novel acceleration algorithm for geometric template matching is proposed based on the Cauchy-Schwartz inequality (C-S inequality) in this paper. The proposed approach is validated under a scanning electron microscope (SEM), and its effectiveness is demonstrated. In this approach, the object shape features are represented as column vectors with complex elements. Then, a threshold is determined to exclude the sliding windows without any matched objects. Finally, surface fitting is employed to obtain the subpixel positions of the targets, and least squares adjustment is utilized to fine-tune the obtained results. The experimental results demonstrate that the proposed method can significantly reduce the matching time by 59% - 96% compared with the traditional shape-based method. Furthermore, the strong robustness and high accuracy are verified under different disturbances. Additionally, the approach is shown to be robust and accurate under different types of disturbances, thus confirming its suitability for real-time targets tracking on SEM with high accuracy.

This paper deals with problems of surface object recognition in urban environmental conditions in infrared sequences using the template matching technique. Target recognition is always necessary in the selection and tracking the defined objects. The objects feature is changed in dependence on time and space condition such as weather condition, urban environment, etc. In this article, we proposed a method for targets recognition using template matching technique in infrared (IR) video sequences. A user simply chooses the given object at some point during detection. On the basis of feature of selected object, the algorithm employed the template matching techniques to find the object. The designed algorithm was tested in program MATLAB and MATLAB - SIMULINK. The paper will deal with a new algorithm using template matching. The idea of the paper will approach to reduce processing time and increase precision for object detection and selection in using template matching technique. There are many similarities criteria for template matching that are sum of absolute differences (SAD), locally scaled sum of absolute difference (LSAD), zero - mean sum of absolute differences (ZSAD), sum of squared differences (SSD), etc. In this paper we have employed the similarity criteria normalized cross correlation. The cross correlation is not invariant to changes in image intensity such as lighting conditions, and the range of correlation coefficient is dependent on the size of the feature, while we can normalize for the effect of changing intensity and template size by using normalized cross correlation. The basic principle of the algorithm is based on the assumption the object is selected at time t with the center of mass T and we also know the greatest relative velocity between the camera and the target ν ct . Thus, the greatest distance which the target can move after the interval Δt will be limited in the circle with the center T at time t and radius R. That means the size of compared image will be smaller than the size of original frame in infrared video sequence. Therefore, the computational time will be faster. In order to reduce the unnecessary computational time of the template matching technique in whole frame, we need to keep the requirement of the selected maximum size of the target is smaller than a half of the size of the frame. It means that the maximum size of the image ƒ(x, y) is equal to the size of the frame in video sequence.

Remote sensing data includes images captured by satellite-based sensors that encode geospatial information. Image registration, an essential preprocessing step, involves aligning satellite images with geometrically precise reference images. Co-registering multi-temporal datasets to reference images is a prerequisite for reliable spatiotemporal analysis and related applications. In this paper, we present a novel satellite image registration method using Spatial Transformer Networks (STNs), where a localization network learns affine transformation parameters between the given input and reference image, and a grid generator-sampler produces the transformed image. The proposed model is capable of learning complex features/patterns despite the spatiotemporal changes. We evaluate our model on a custom multi-temporal ResourceSAT-2/2A LISS-3 dataset using two training approaches. Approach 1 involves applying the loss function after the grid generator-sampler stage, simulating practical scenarios where transformation parameters are unknown. Approach 2 applies the loss function after the parameter retrieval stage, leveraging known parameters from our custom dataset. In Approach 1, we evaluated four distinct models using various hyperparameter configurations and conducted a comparative analysis. The results, based on normalized cross-correlation (NCC) loss computed between the transformed and reference images using the Adam optimizer, were promising. In Approach 2, the NCC loss with the Adam optimizer was leveraged, but here it measures the error between the predicted and the known affine parameters. Our proposed method yields sub-pixel-level accurate image registration. We performed a comparative analysis of the results from Approaches 1 and 2 against existing methods to evaluate their relative performance.

A novel approved algorithm based on template matching is proposed. This algorithm is used to predict the movement of the target and track the target. By utilizing the DESO(Differential Extended State Observer) filter, it is possible to obtain the estimate position of the target in the current image. The template matching algorithm is used to search target around the estimation position. The simulation results of the algorithm are presented. The proposed algorithm demonstrates better performance characteristics than traditional template matching algorithm. Experimental results also demonstrate the effectiveness of this method. Through the simulation experiment, it can be seen that regardless of the goal of uniform motion or motion with variable velocity, DESO motion prediction based on the template matching algorithm can accurately predict the trajectory of the target, and does not depend on the target motion model. The use of DESO is simple, the transition time is short and the tracking performance is satisfying.

In this work, we propose a new approach for three-dimensional registration of MR fractional anisotropy images with T1-weighted anatomy images of human brain. From the clinical point of view, this accurate coregistration allows precise detection of nerve fibers that is essential in neuroscience. A template matching algorithm combined with normalized cross-correlation was used for this registration task. To show the suitability of the proposed method, it was compared with the normalized mutual information-based B-spline registration provided by the Elastix software library, considered a reference method. We also propose a general framework for the evaluation of robustness and reliability of both registration methods. Both registration methods were tested by four evaluation criteria on a dataset consisting of 74 healthy subjects. The template matching algorithm has shown more reliable results than the reference method in registration of the MR fractional anisotropy and T1 anatomy image data. Significant differences were observed in the regions splenium of corpus callosum and genu of corpus callosum, considered very important areas of brain connectivity. We demonstrate that, in this registration task, the currently used mutual information-based parametric registration can be replaced by more accurate local template matching utilizing the normalized cross-correlation similarity measure.

As oil palm is cultivated in large-scale plantations, prior and post-planting operations are laborious and time expensive, and also manual implementation of some of these operations often results in incorrect measurements and information. Successful oil palm management in prior and post-planting operations requires effective techniques to collect precise information. Unmanned aerial vehicle (UAV) imagery is a low-cost alternative to field-based assessment but requires the development of methods to easily and accurately extract the required information. The individual oil palm tree detection and height assessment are important and labor-intensive tasks for large-scale plantations where trees taller than 15 m are being replanted due to the high cost of harvesting and low yield. Therefore, in this chapter we demonstrate a general work flow for individual oil palm detection and height assessment using UAV imagery. We explain local maximum (LM) and template matching (TM) techniques as two commonly used approaches in individual tree detection. The accuracy of each method was evaluated on 20 randomly selected plots on UAV image with recall, precision, and F-score method. The F-score for LM method is higher than TM methods which, respectively, are 0.83 and 0.60. From 17,252 oil palm trees in the 20 plots, LM algorithm could detect 1395 (almost 80%) and TM algorithm, 967 (almost 55%). Three hundred fifty-seven and 785 trees have been missed, respectively, in LM and TM approaches. In both cases, background vegetation incorrectly has been labeled as oil palm tree, where 141 and 322 objects have been falsely detected as oil palm tree in LM and TM approaches. LM worked better in almost all of the plots; however the performance decreased in highly dense area. Contradictory to TM approach, in LM approach, shadows did not affect the performance as it was reflected in precision value. In densely cultivated plots due to leaves overlapping of neighboring trees, algorithm failed, and also in sparsely cultivated plots, shadows caused some commission errors. Inherited distortion of UAV image also caused some omission errors in TM approach. Individual detected tree with LM algorithm was used in the next part for oil palm height estimation overlaying with canopy height model.