1,724,312 research outputs found
Efficient generic calibration method for general cameras with single centre of projection
Get PDFGeneric camera calibration is a non-parametric calibration technique that is applicable to any type of vision sensor. However, the standard generic calibration method was developed with the goal of generality and it is therefore sub-optimal for the common case of cameras with a single centre of projection (e.g. pinhole, fisheye, hyperboloidal catadioptric). This paper proposes novel improvements to the standard generic calibration method for central cameras that reduce its complexity, and improve its accuracy and robustness. Improvements are achieved by taking advantage of the geometric constraints resulting from a single centre of projection. Input data for the algorithm is acquired using active grids, the performance of which is characterised. A new linear estimation stage to the generic algorithm is proposed incorporating classical pinhole calibration techniques, and it is shown to be significantly more accurate than the linear estimation stage of the standard method. A linear method for pose estimation is also proposed and evaluated against the existing polynomial method. Distortion correction and motion reconstruction experiments are conducted with real data for a hyperboloidal catadioptric sensor for both the standard and proposed methods. Results show the accuracy and robustness of the proposed method to be superior to those of the standard method
Weakly Supervised Learning of Objects, Attributes and their Associations
No full textAbstract. When humans describe images they tend to use combinations of nouns and adjectives, corresponding to objects and their associated attributes respectively. To generate such a description automatically, one needs to model objects, attributes and their associations. Conventional methods require strong annotation of object and attribute locations, making them less scalable. In this paper, we model objectattribute associations from weakly labelled images, such as those widely available on media sharing sites (e.g. Flickr), where only image-level labels (either object or attributes) are given, without their locations and associations. This is achieved by introducing a novel weakly supervised non-parametric Bayesian model. Once learned, given a new image, our model can describe the image, including objects, attributes and their associations, as well as their locations and segmentation. Extensive experiments on benchmark datasets demonstrate that our weakly supervised model performs at par with strongly supervised models on tasks such as image description and retrieval based on object-attribute associations
Improving Bag-of-Words model with spatial information
No full textBag-of-Words (BOW) models have recently become popular for the task of object recognition, owing to their good performance and simplicity. Much work has been proposed over the years to improve the BOW model, where the Spatial Pyramid Matching technique is the most notable. In this work, we propose three novel techniques to capture more re_ned spatial information between image features than that provided by the Spatial Pyramids. Our techniques demonstrate a performance gain over the Spatial Pyramid representation of the BOW model
Single Image Defocus Deblurring Using Kernel-Sharing Parallel Atrous Convolutions
No full textThis paper proposes a novel deep learning approach for single image defocus deblurring based on inverse kernels. In a defocused image, the blur shapes are similar among pixels although the blur sizes can spatially vary. To utilize the property with inverse kernels, we exploit the observation that when only the size of a defocus blur changes while keeping the shape, the shape of the corresponding inverse kernel remains the same and only the scale changes. Based on the observation, we propose a kernel-sharing parallel atrous convolutional (KPAC) block specifically designed by incorporating the property of inverse kernels for single image defocus deblurring. To effectively simulate the invariant shapes of inverse kernels with different scales, KPAC shares the same convolutional weights among multiple atrous convolution layers. To efficiently simulate the varying scales of inverse kernels, KPAC consists of only a few atrous convolution layers with different dilations and learns per-pixel scale attentions to aggregate the outputs of the layers. KPAC also utilizes the shape attention to combine the outputs of multiple convolution filters in each atrous convolution layer, to deal with defocus blur with a slightly varying shape. We demonstrate that our approach achieves state-of-the-art performance with a much smaller number of parameters than previous methods1
Convergence Analysis of MAP based Blur Kernel Estimation
No full textOne popular approach for blind deconvolution is to formulate a maximum a posteriori (MAP) problem with sparsity priors on the gradients of the latent image, and then alternatingly estimate the blur kernel and the latent image. While several successful MAP based methods have been proposed, there has been much controversy and confusion about their convergence, because sparsity priors have been shown to prefer blurry images to sharp natural images. In this paper, we revisit this problem and provide an analysis on the convergence of MAP based approaches. We first introduce a slight modification to a conventional joint energy function for blind deconvolution. The reformulated energy function yields the same alternating estimation process, but more clearly reveals how blind deconvolution works.
We then show the energy function can actually favor the right solution instead of the no-blur solution under certain conditions, which explains the success of previous MAP based approaches. The reformulated energy function and our conditions for the convergence also provide a way to compare the qualities of different blur kernels, and we demonstrate its applicability to automatic blur kernel size selection, blur kernel estimation using light streaks, and defocus estimation.1
Robust metric structure from motion for an extended sequence with outliers and missing data
No full text[[abstract]]©2005 ICCV BenCOS-In this paper, we propose a robust metric structure from motion (SfM) algorithm for an extended sequence with outliers and missing data. There are three main contributions in the proposed SfM algorithm. The first is a novel jury-based preemptive LMedS procedure to achieve efficient outlier detection. The second contribution is a new iterative two-step scheme that consists of robust estimation techniques for projective structure from motion. The third contribution is a novel algorithm for robust metric upgrade by applying the M-estimator to the traditional linear constraints for metric upgrade. In addition, comparisons of the proposed algorithm with some previous methods through experiments on simulated data are shown to demonstrate the efficiency and robustness of the proposed algorithm[[department]]資訊工程學
Descriptive temporal template features for visual motion recognition
No full textCopyright © 2009 Elsevier B.V. All rights reserved.In this paper, a human action recognition system is proposed. The system is based on new, descriptive ‘temporal template’ features in order to achieve high-speed recognition in real-time, embedded applications. The limitations of the well-known ‘Motion History Image’ (MHI) temporal template are addressed and a new ‘Motion History Histogram’ (MHH) feature is proposed to capture more motion information in the video. MHH not only provides rich motion information, but also remains computationally inexpensive. To further improve classification performance, we combine both MHI and MHH into a low dimensional feature vector which is processed by a support vector machine (SVM). Experimental results show that our new representation can achieve a significant improvement in the performance of human action recognition over existing comparable methods, which use 2D temporal template based representations
ClassCut for Unsupervised Class Segmentation
Get PDFWe propose a novel method for unsupervised class segmentation on a set of images. It alternates between segmenting object instances and learning a class model. The method is based on a segmentation energy defined over all images at the same time, which can be optimized efficiently by techniques used before in interactive segmentation. Over iterations, our method progressively learns a class model by integrating observations over all images. In addition to appearance, this model captures the location and shape of the class with respect to an automatically determined coordinate frame common across images. This frame allows us to build stronger shape and location models, similar to those used in object class detection. Our method is inspired by interactive segmentation methods [1], but it is fully automatic and learns models characteristic for the object class rather than specific to one particular object/image. We experimentally demonstrate on the Caltech4, Caltech101, and Weizmann horses datasets that our method (a) transfers class knowledge across images and this improves results compared to segmenting every image independently; (b) outperforms Grabcut [1] for the task of unsupervised segmentation; (c) offers competitive performance compared to the state-of-the-art in unsupervised segmentation and in particular it outperforms the topic model [2]
Skin segmentation for robust face image analysis
Get PDFThis thesis presents novel techniques to address the challenge of outlier detection and removal in the context of face analysis from photographs. Given a face image, under arbitrary scene conditions, our goal is to automatically compute a binary map that indicates the locations of facial occlusions, such as hairstyle, beard, clothing or glasses, and other atypical elements. The motivation is that this information can help other face processing methods, which do not tackle this problem on their own, to improve their results with minimal algorithmic adjustments. The 3D Morphable Model is a good example for such a method, and serves as testbed for our finding.
Usually outliers are difficult to capture. By definition they represent unpredictable deviations from facial appearance, which elude a systematical analysis. The problem is, that outliers impair a face description by perturbing extracted features. This can lead to wrong classifications or otherwise defective outputs. Therefore, in the face recognition literature, several methods have been devised to deal with this phenomenon. However, these solutions are neither comparable to our approach, nor applicable to our target applications, as they are often suited to a specific feature representation and not comprehensive.
We address the outlier problem, for the first time, as a classical segmentation task. The main contribution of our work is an algorithm, which determines the location of outliers on a pixel scale, by partitioning a face image into skin and "non-skin" regions. The algorithm is designed to work completely automatic and, unlike conventional skin detection techniques, it does not depend on color input. The latter is accomplished by means of a novel low-level texture analysis procedure, which comprises an illumination compensation step and a subsequent matching of image regions with respect to a given sample of skin texture. The resulting texture features are segmented with a customized version of the supervised GrabCut method. In order to facilitate automation, we incorporate structural knowledge on faces from the 3D Morphable Model. It allows us to mark specific facial areas, which are utilized as skin samples as well as to inizialized the actual segmentation routine.
We demonstrate the significance of the skin segmentation on three applications. First, it serves as main component to create an outlier map, that works in combination with a slightly modified fitting algorithm, to greatly improve the visual quality of 3D Morphable Model reconstructions. The second application extends this capability and reuses the image content, associated with the outliers, to realize a high level photo manipulation, called Face Exchange. The aim here is to substitute faces between different images, without affecting the rest of the scene. The last contribution represents a novel approach to face recognition. We localize prominent irregularities in facial skin, particularly moles, in order to use their characteristic configuration within a face for identification. For this task the skin segments are of utmost importance, to ensure high detection accuracy, and expressiveness of the extracted features
- …
