By Topic

Computer Vision, IET

Issue 3 • Date September 2010

Filter Results

Displaying Results 1 - 7 of 7
  • Height estimation from monocular image sequences using dynamic programming with explicit occlusions

    Page(s): 149 - 161
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1032 KB)  

    In this study, the authors propose a novel algorithm to estimate the heights of objects from monocular aerial images taken from mobile platforms such as unmanned aerial vehicles and small airplanes. Sequential images captured by a single camera mounted on a mobile platform contain 3D information of objects. In this study, the authors propose to use illumination normalisation to reduce illumination variations and to use at least two objects with known distances to accurately estimate the camera focal length. The authors also propose a novel stereo matching algorithm using dynamic programming with explicit occlusion modelling to recover depth information in occluded regions and to preserve depth discontinuity. As a result, the authors are able to reliably estimate the heights of objects in or close to power line corridors. Our experiments show that the proposed algorithm can estimate the heights of trees and power poles from aerial images with average errors of 1.8 and 1.1 m, respectively, when the flight height is in the range between 230 and 280 m above ground level. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Statistical interpretation of non-local means

    Page(s): 162 - 172
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (706 KB)  

    Noise filtering is a common step in image processing, and is particularly effective in improving the subjective quality of images. A large number of techniques have been developed, many of which concentrate on the problem of removing noise without damaging small structures such as edges. One recent approach that demonstrates empirical merit is the non-local means (NLM) algorithm. However, in order to use noise filtering algorithms in quantitative or clinical image analysis tasks an understanding of their behaviour that goes beyond subjective appearance must be developed. The purpose of this study is to investigate the statistical basis of NLM in order to attempt to understand the conditions required for its use. The theory is illustrated on synthetic data and clinical magnetic resonance images of the brain. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Preserving global and local information - a combined approach for recognising face images

    Page(s): 173 - 182
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (393 KB)  

    Face recognition can significantly impact authentication, monitoring and indexing applications. Much research on face recognition using global and local information has been done earlier. By using global feature preservation techniques like principal component analysis (PCA) and linear discriminant analysis (LDA), the authors can effectively preserve only the Euclidean structure of face space that suffers lack of local features, but which may play a major role in some applications. On the other hand, the local feature preservation technique namely locality preserving projections (LPP) preserves local information and obtains a face subspace that best detects the essential face manifold structure; however, it also suffers loss in global features which may also be important in some of the applications. A new combined approach for recognising faces that integrates the advantages of the global feature extraction technique LDA and the local feature extraction technique LPP has been introduced here. Xiaofei He et al. in their work used PCA to extract similarity features from a given set of images followed by LPP. But in the proposed method, the authors use LDA (instead of PCA) to extract discriminating features that yields improved facial image recognition results. This has been verified by making a fair comparison with the existing methods. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Aspect coherence for graph-based semantic image labelling

    Page(s): 183 - 194
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (712 KB)  

    In image semantic segmentation a semantic category label is associated to each image pixel. This classification problem is characterised by pixel dependencies at different scales. On a small-scale pixel correlation is related to object instance sharing, whereas on a middle- and large scale to category co-presence and relative location constraints. The contribution of this study is two-fold. First, the authors present a framework that jointly learns category appearances and pixel dependencies at different scales. Small-scale dependencies are accounted by clustering pixels into larger patches via image oversegmentation. To tackle middle-scale dependencies a conditional random field (CRF) is built over the patches. A novel strategy to exploit local patch aspect coherence is used to impose an optimised structure in the graph to have exact and efficient inference. The second contribution is a method to account for full patch neighbourhoods without introducing loops in the graphical structures. `Weak neighbours` are introduced, which are patches connected in the image but not in the inference graph. They are pre-classified according to their visual appearance and their category distribution probability is then used in the CRF inference step. Experimental evidence of the validity of the method shows improvements in comparison to other works in the field. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Face recognition using enhanced linear discriminant analysis

    Page(s): 195 - 208
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1303 KB)  

    There are two fundamental problems with the linear discriminant analysis (LDA) for face recognition. First one is LDA is not stable because of the small training sample size problem. The other is that it would collapse the data samples of different classes into one single cluster when the class distributions are multimodal. An enhanced LDA method is proposed to overcome these two problems. The between- and within-class scatters are reformulated by introducing two different weighted matrices in respective. The enhanced Fisher criterion is then presented, which can preserve the local structure of different class in the reduced subspace. Moreover, maximum margin criterion is adopted to avoid the singularity problem of the within-class scatter matrix. Extensive experiments show encouraging recognition performance of the proposed algorithm. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multimodal biometric method based on vein and geometry of a single finger

    Page(s): 209 - 217
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (673 KB)  

    There are limits to a single biometric observation such as variation in an individual biometric feature due to the condition of a sensor, the health condition of a human, illumination variation and so on. To overcome such limitations, the authors propose a new multimodal biometric approach integrating finger vein recognition and finger geometry recognition at the score level. The method presents three advantages compared to previous works: (i) the proposed multimodal biometric system can be constructed as a tiny device, which uses a finger vein and finger geometry features acquired from a single finger; (ii) the proposed finger geometry recognition, based on Fourier descriptors, is robust to the translation and rotation of a finger; and (iii) the authors obtained better recognition accuracy using the score-level fusion method based on a support vector machine than by other score-level fusion methods such as the MAX, MIN and SUM rules. The results showed that the equal error rate of the proposed method decreased by as much as 1.089 and 1.627% compared with finger vein recognition and finger geometry recognition methods, respectively. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Modelling visual saliency using degree centrality

    Page(s): 218 - 229
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (961 KB)  

    Visual attention is an indispensable component of complex vision tasks. A multi-scale, complex network-based approach for determining visual saliency is described. It uses degree centrality (conceptually and computationally the simplest among all the centrality measures) over a network of image regions to form a saliency map. The regions used in the network are multiscale in nature with scale selected automatically. Experimental evaluation establishes the superiority of the method over existing saliency methods, even in noisy environments. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.

Aims & Scope

IET Computer Vision seeks original research papers in a wide range of areas of computer vision. The vision of the journal is to publish the highest quality research work that is relevant and topical to the field, but not forgetting those works that aim to introduce new horizons and set the agenda for future avenues of research in Computer Vision.

Full Aims & Scope