No abstract available.

A saliency map specifies the relevance of locations in an image and good methods exist today for recovering it automatically in a bottom-up fashion. This paper addresses the problem of finding organization in an actual saliency map of everyday images through a grouping field in order to elicit structure that can be used for a variety of purposes. After calculating the saliency using a version of the Itti et al. model, we retain only the maxima of the map as features. Then we calculate an initial grouping field by convolving the salient locations with orientation-selective grouping kernels. Finally we iteratively reinforce the field to enhance perceptually salient components.

We describe a method to enable the selection of specific text regions with a hand-held camera by means of projecting a structured light pointer on the document. The user indicates the text required by dragging the laser pointer over it while a sequence of images is captured. By tracking the motion of the camera over the document and matching the trajectory to the actual text, the system is able to precisely determine the text portion the user intended to capture. By using this selection method along with a text-processing pipeline and OCR, a general purpose hand-held device (such as a PDA or mobile phone) with a camera could be used as effectively as single-purpose pen scanning devices. We present our results showing successful capture and extraction of text.

This paper presents a fast and extremely robust feature-based method for planar registration of partly overlapping images that uses a twostage robust tting approach comprising a fast estimation of a transformation hypothesis (that we show is highly likely to be correct) followed by a con rmation and re nement stage. The method is particularly suited for automatic stitching of oversize documents scanned in two or more parts. We show simulations, also supported by practical experiments, that prove both the robustness and computational eAEciency of the approach.

In the past few years we have witnessed the migration of cheap imaging devices into portable and mobile computing appliances such as PDAs and mobile phones. As these devices become ever more powerful in terms of processor speed and memory, new exciting applications and uses are being developed. A particularly useful one is the casual capture of text images for faxing, note taking, OCR, etcetera. This paper describes the image processing pipeline used to enhance images of text captured by a hand-held low-resolution camera, and a fast text extraction method. The main advantages of the approach are its inherent lightweight structure, speed and relative robustness under poor lighting and focus conditions. The computational efficiency (and a careful implementation) of the approach has allowed its deployment in an interactive-time text capture and foreign-text translation demo on a PDA with a VGA camera attachment.

This paper presents a fast and extremely robust feature-based method for planar registration of partly overlapping images that uses a two-stage robust fitting approach comprising a fast estimation of a transformation hypothesis (that we show is highly likely to be correct) followed by a confirmation and refinement stage. The method is particularly suited for automatic stitching of oversize documents scanned in two or more parts. We show simulations, also supported by practical experiments, that prove both the robustness and computational efficiency of the approach.

The work presented is concerned with the reconstruction of the original undistorted image of a curled document or book when captured face-up by a camera with the support of sparse depth measurements. A novel method based on physically modeling paper deformation with an applicable surface is proposed and a relaxation algorithm is described that allows us to fit it to noisy data and then flatten it in order to produce the final undistorted image. The promising results obtained confirm that the use of applicable surfaces is the right way to address this problem.

The article deals with the recovery of illusory linear clues from perspectively skewed documents, with the purpose of using them for rectification. The computational approach proposed implements the perceptual organization principles implicitly used in textual layouts. The numerous examples provided show that the method is robust and viewpoint and scale invariant.

This work is concerned with the reconstruction of the flat state of a curled document or book page when captured face up by a camera. It is reconstructed with the support of sparse depth measurements of its surface. A novel method based on physical modeling of paper deformation with an applicable surface is proposed and a relaxation algorithm is described that allows us to fit this model and unroll it to a plane so as to produce the undistorted document. Some very promising results are shown that confirm that the use of applicable surfaces is the right way to address this problem.

This paper addresses the optimization of pseudo-random planar point patterns for invariant-based identification or indexing. This is a novel problem and is formulated here as the maximization of the spacing of all the invariants when considered as points in a space. The task is of formidable complexity and a stochastic approximation strategy is proposed that yields interesting results.

This work presents a new efficient method for fitting ellipses to scattered data. Previous algorithms either fitted general conics or were computationally expensive. By minimizing the algebraic distance subject to the constraint 4ac-b/sup 2/=1, the new method incorporates the ellipticity constraint into the normalization factor. The proposed method combines several advantages: It is ellipse-specific, so that even bad data will always return an ellipse. It can be solved naturally by a generalized eigensystem. It is extremely robust, efficient, and easy to implement.

Abstract This paper presents a novel automatic method for view synthesis (or image transfer) from a triplet of uncalibrated images based on trinocular edge matching followed by transfer by interpolation, occlusion detection and correction and finally rendering. The edge-based technique proposed here is of general practical relevance because it does not rely upon dense correspondences and is computationally efficient. Applications range from immersive media and teleconferencing, image interpolation for fast rendering and compression.

This paper presents a novel automatic method for view synthesis (or image transfer) from a triplet of uncalibrated images based on trinocular edge matching followed by transfer by interpolation, occlusion detection and correction and finally rendering. The edge-based technique proposed here is of general practical relevance because it overcomes most of the problems encountered in other approaches that either rely upon dense correspondence, work in projective space or need explicit camera calibration. Applications range from immersive media and teleconferencing, image interpolation for fast rendering and compression.

This paper proposes a new algorithm for matching point features across pairs of images. Despite the well-known combinatorial complexity of the problem, this work shows that an acceptably good solution can be obtained directly by singular value decomposition of an appropriate correspondence strength matrix. The approach draws from the method proposed previously but, besides suggesting its usefulness for stereo matching, in this work a correlation-weighted proximity function is used as correspondence strength to specifically cater for real images.

In the context of part segmentation from 2D edge images, this paper presents some interesting results with a novel method that addresses the problem of filtering a redundant set of part hypotheses that retains only those that are likely to correspond to actual parts. In the proposed method, supporting evidence for hypotheses competes in a minimum description-length (MDL) framework to select part hypotheses that most economically represent supporting edges in the 'language' of generic parts.

This paper addresses the following problem: How can we make a complicated mathematical shape model simpler while keeping a comparable level of representational power? The proposed solution is to use the original model itself -- which represents a class of shapes -- to train a Point Distribution Model. In this paper the idea is applied to the case of deformable superellipses.

Ellipse fitting is one of the classic problems of pattern recognition and has been subject to considerable attention because of its many applications. This article presents the first direct method for specifically fitting ellipses in the least squares sense. Previous approaches used either generic conic fitting or relied on iterative methods to recover elliptic solutions. The proposed method is (i) ellipse-specific, (ii) directly solved by a generalised eigen-system, (iii) has a desirable low-eccentricity bias, and (iv) is robust to noise. We provide a theoretical demonstration, several examples and the Matlab coding of the algorithm.

This paper presents a new efficient method for fitting ellipses to scattered data. Previous algorithms either fitted general conics or were computationally expensive. By minimizing the algebraic distance subject to the constraint 4ac-b/sup 2/=1 the new method incorporates the ellipticity constraint into the normalization factor. The new method combines several advantages: 1) it is ellipse-specific so that even bad data will always return an ellipse; 2) it can be solved naturally by a generalized eigensystem, and 3) it is extremely robust, efficient and easy to implement. We compare the proposed method to other approaches and show its robustness on several examples in which other nonellipse-specific approaches would fail or require computationally expensive iterative refinements.

This work proposes a methodology for the analysis of the uncertainty in the localization of objects when considering uncertain image data, camera and object geometry parameters. The uncertainty is propagated through an extended static Kalman filter initialized with the parameters used for the localization and updated with new matched features obtained by back-projecting onto the image. At the end of the process, a better estimate of the object pose with its uncertainty is given along with a new estimate of the used uncertain object features and the camera parameters. The methodology is now in use in an object localization system. 1 Introduction The role of uncertainty is very important where a measure of the position and orientation of a modeled object in space is required. Besides providing information on the reliability of the data, the reason for using the uncertainty of the sensory data is for improving the estimate of the parameters we want to measure. This improvement is obtaine...