Point set registration is a key component in many computer vision tasks. This paper proposes a point set registration algorithm based on information geometry. Point sets to be registration are converting to the statistical manifolds by Gaussian mixture model. The component of mixture model represents the dimension of statistical manifold and point set is a point on manifold. Through conversion, point set registration is reformulated as searching the shortest path between two manifold and we can use the em algorithm which defined by information geometry to get the optimization solution. Experimental results show that the proposed algorithm is robust to noise and outliers, and achieved very good accuracy.
Swarm intelligence-based image thresholding segmentation algorithms are playing an important role in the research field of image segmentation. In this paper, we briefly introduce the theories of four existing image segmentation algorithms based on swarm intelligence including fish swarm algorithm, artificial bee colony, bacteria foraging algorithm and particle swarm optimization. Then some image benchmarks are tested in order to show the differences of the segmentation accuracy, time consumption, convergence and robustness for Salt&Pepper noise and Gaussian noise of these four algorithms. Through these comparisons, this paper gives qualitative analyses for the performance variance of the four algorithms. The conclusions in this paper would give a significant guide for the actual image segmentation.
A speed up technique for the SUSAN edge detector based on random sampling is proposed. Instead of sliding the mask pixel by pixel on an image as the SUSAN edge detector does, the proposed scheme places the mask randomly on pixels to find edges in the image; we hereby name it randomized SUSAN edge detector (R-SUSAN). Specifically, the R-SUSAN edge detector adopts three approaches in the framework of random sampling to accelerate a SUSAN edge detector: procedure integration of response computation and nonmaxima suppression, reduction of unnecessary processing for obvious nonedge pixels, and early termination. Experimental results demonstrate the effectiveness of the proposed method.
Stitching of images is the task of fusing a collection of images with smaller field of view (FOV) to obtain an image with a larger FOV. The need for this procedure is motivated by a variety of applications, especially including the application in the aviation remote sensing field. The paper puts emphasis on studying how to stitch the image set taken by the aviation swaying camera. In order to accomplish this task, we should know how the aviation camera sways. So the camera swaying model is presented firstly. The possible geometrical distortion of the image set can be known from the camera swaying model, and it includes two parts: the rotation and the perspective distortion. Aiming at the rotation distortion, we propose a novel image rotation method that can save the computation time. In this method, we rotate the four edge lines of the image utilizing the traditional rotation method. According to the relative position rotation invariability of any pair of points in an image, the other points except that on the four edge lines then can be rotated. The last step of the algorithm is the stitching which needs not only to search the optimal matching area between neighboring images but also to stitch the image set smoothly. The paper presents an extremum matching method based on the difference image, by which we can find the optimal matching areas between neighboring images in the image set. Through linearly changing the stitching weight of overlap areas, the image set can be stitched smoothly. The theoretical analysis and the experimental results have demonstrated the feasibility and validity of this stitching algorithm for the remote sensing images from the aviation swaying camera.
According to the imaging mechanism of the IR image, the histogram of the IR image, whose background is composed of the sea area and the sky area, usually has two apices and one vale. Based on this character of the histogram, the paper proposes a background suppression method, which combines the sea-sky area segmentation with the median filtering. Then, in order to reduce random noises in the image whose background has been suppressed, the local threshold-a digitizing method-has been showed in the paper. In this method, the image whose background has been suppressed is divided into some blocks, and the different block has the different threshold. The target motion is continuous and the noise motion is random, so the paper presents a sequential target detection method on the centroid track accumulation. A large number of experiments show that the methods presented in the paper can exactly detect the moving small target in the IR clutter background containing sea and sky areas.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.