Electronic Letters on Computer Vision and Image Analysis (ELCVIA - Universitat Autònoma de Barcelona)
Not a member yet
343 research outputs found
Sort by
Two View Line-Based Motion and Structure Estimation for Planar Scenes
Get PDFWe present an algorithm for reconstruction of piece-wise planar scenes from only two views and based on minimum line correspondences. We first recover camera rotation by matching vanishing points based on the methods already exist in the literature and then recover the camera translation by searching among a family of hypothesized planes passing through one line. Unlike algorithms based on line segments, the presented algorithm does not require an overlap between two line segments or more that one line correspon- dence across more than two views to recover the translation and achieves the goal by exploiting photometric constraints of the surface around the line. Experimental results on real images prove the functionality of the algorithm
A Font Search Engine for Large Font Databases
Get PDFA search engine for font recognition is presented and evaluated. The intended usage is the search in very large font databases. The input to the search engine is an image of a text line, and the output is the name of the font used when rendering the text. After pre-processing and segmentation of the input image, a local approach is used, where features are calculated for individual characters. The method is based on eigenimages calculated from edge filtered character images, which enables compact feature vectors that can be computed rapidly. In this study the database contains 2763 different fonts for the English alphabet. To resemble a real life situation, the proposed method is evaluated with printed and scanned text lines and character images. Our evaluation shows that for 99.1% of the queries, the correct font name can be found within the five best matches
A reduced domain pool based on DCT for a fast fractal image encoding
Get PDFFractal image compression is time consuming due to the search of the matching between range and domain blocks. In order to improve this compression method, we propose firstly, in this paper, a fast method for reducing the computational complexity of fractal encoding by reducing the size of the domain pool. This reduction is based on the lowest horizontal and vertical DCT coefficients of domain blocks. The experimental results on the test images show that the proposed method reduce the time computation and reach a high speedup factor without decreasing the image quality. Secondly, we combine our method to the AP2D approach which uses two domain pools in two steps of encoding. A more reduction of encoding time is obtained without decreasing the image quality
Combining Total Variation and Nonlocal Means Regularization for Edge Preserving Image Deconvolution
Get PDFWe propose a new edge preserving image deconvolution model by combining total variation and nonlocal means regularization. Natural images exhibit an high degree of redundancy. Using this redundancy, the nonlocal means regularization strategy is a good technique for detail preserving image restoration. In order to further improve the visual quality of the nonlocal means based algorithm, total variation is introduced to the model to better preserve edges. Then an efficient alternating minimization procedure is used to solve the model. Numerical experiments illustrate the effectiveness of the proposed algorithm
Implementation of Max Principle with PCA in image fusion for Surveillance and Navigation Application
Get PDFImage fusion is the combination of two or more different images by using suitable algorithms to form an output image. It provides a useful tool to integrate multiple images into a composite image. In this paper, we present an approach that uses the principle component transform along with the selection of maximum pixel intensity to perform pixel level fusion. The entropy, mutual information and the universal index based measure are used to evaluate the performance of this fusion algorithm. Keywords: Wavelet transformation, pixel level image fusion, PC
Simple fish-eye calibration method with accuracy evaluation
Get PDFIn this paper, a simple fish-eye radial distortion calibration procedure is described. This method avoids costly minimisation and optimisation algorithms, and is based on trivial concentricity of three extracted points. The results show that this simplicity is at the expense of increased deviation of results (and thus increased error). However, this deviation can be reduced significantly by the use of simple averaging, such that it is only marginally greater than the current state-of-the-art
Formalization of the General Video Temporal Synchronization Problem
Get PDFIn this work, we present a theoretical formalization of the temporal synchronization problem and a method to temporally synchronize multiple stationary video cameras with overlapping views of the same scene. The method uses a two stage approach that first approximates the synchronization by tracking moving objects and identifying curvature points. The method then proceeds to refine the estimate using a consensus based matching heuristic to find frames that best agree with the pre-computed camera geometries from stationary background image features. By using the fundamental matrix and the trifocal tensor in the second refinement step, we improve the estimation of the first step and handle a broader more generic range of input scenarios and camera conditions. The method is relatively simple compared to current techniques and is no harder than feature tracking in stage one and computing accurate geometries in stage two. We also provide a robust method to assist synchronization in the presence of inaccurate geometry computation, and a theoretical limit on the accuracy that can be expected from any synchronization system
A Performance Evaluation of Exact and Approximate Match Kernels for Object Recognition
Get PDFLocal features have repeatedly shown their effectiveness for object recognition during the last years, and they have consequently become the preferred descriptor for this type of problems. The solution of the correspondence problem is traditionally approached with exact or approximate techniques. In this paper we are interested in methods that solve the correspondence problem via the definition of a kernel function that makes it possible to use local features as input to a support vector machine. We single out the match kernel, an exact approach, and the pyramid match kernel, that uses instead an approximate strategy. We present a thorough experimental evaluation of the two methods on three different databases. Results show that the exact method performs consistently better than the approximate one, especially for the object identification task, when training on a decreasing number of images. Based on this findings and on the computational cost of each approach, we suggest some criteria for choosing between the two kernels given the application at hand.
A robust multi-feature cut detection algorithm for video segmentation
Get PDFVideo segmentation is the first task in almost all video analysis applications. It consists in identifying the boundaries of the meaningful video units (shots). Without a doubt, cuts are the most common among production effects that characterize the shot boundaries. In this paper we propose an algorithm for cut detection exploiting an innovative, robust frame difference measure. The measure is based on a combination of different visual features. To improve the precision of the cut detection algorithm, a temporal pattern analysis model, and a flashes removal are also proposed.Experimental results to prove the effectiveness of the proposed measure coupled with the temporal pattern analysis model on very heterogeneous and complex sets of videos are critically reported
Color Image Segmentation using Fast Fuzzy C-Means Algorithm
No full textThis paper proposes modified FCM (Fuzzy C-means) approach to color image segmentation using JND (Just Noticeable Difference) histogram. Histogram of the given color image is computed using JND color model. This samples the color space so that just enough number of histogram bins are obtained on each axis without compromising the visual image content. The number of histogram bins are further reduced using agglomeration. This agglomerated histogram yields the estimation of number of clusters, cluster seeds and the initial fuzzy partition for FCM algorithm. This is a novell approach to estimate the input parameters for FCM algorithm. Then the modified FCM algorithm is proposed that works on histogram bins as data elements instead of individual pixels. This significantly reduces the time complexity of FCM algorithm. To verify the effectiveness of the proposed image segmentation approach, its performance is evaluated on Berkeley Segmentation Database(BSD). Two significant criterias namely PSNR and PRI (Probabilistic Rand Index) are used to evaluate the performance. Results show that the proposed algorithm applied to the JND histogram bins converges much faster and also gives better results than conventional FCM algorithm in terms of PSNR and PRI