Catálogo de publicaciones - libros

This paper addresses the problem of estimating the 3D rigid pose of a CT volume of an object from its 2D X-ray projections. We use maximization of mutual information, an accurate similarity measure for multi-modal and mono-modal image registration tasks. However, it is known that the standard mutual information measure only takes intensity values into account without considering spatial information and its robustness is questionable. In this paper, instead of directly maximizing mutual information, we propose to use a variational approximation derived from the Kullback-Leibler bound. Spatial information is then incorporated into this variational approximation using a Gibbs random field model. The newly derived similarity measure has a least-squares form and can be effectively minimized by a multi-resolution Levenberg-Marquardt optimizer. Experimental results are presented on X-ray and CT datasets of a plastic phantom and a cadaveric spine segment.

Dynamic Programming (DP) matching has been applied to solve distortion in spectral-based fingerprint recognition. However, spectral data is redundant, and its size is huge. PCA could be used to reduce the data size, but leads to loss of topographical information in projected vectors. This allows only inter-vector similarity estimations such as Euclid or Mahalanobis distances, and proves to be inadequate in presence of distortion occurring in finger sweeping with a line sensor. In this paper, we propose a novel two-step PCA to extract compact eigenfeatures amenable to DP matching. The first PCA extracts eigenfeatures of Fourier spectra from each image line. The second extracts eigenfeatures from all lines to form the feature templates. In matching, the feature templates are inversely transformed to line-by-line representations on the first PCA subspace for DP matching. Fingerprint matching experiments demonstrate the effectiveness of our proposed approach in template size reduction and accuracy improvement.

We consider how continuous-time quantum walks can be used for graph matching, both exact and inexact, and measuring graph similarity. Our approach is to simulate the quantum walk on the two graphs in parallel by using an auxiliary graph that incorporates both graphs. The auxiliary graph allows quantum interference to take place between the two walks. Modelling the resultant interference amplitudes, which result from the differences in the two walks, we calculate probabilities for matches between pairs of vertices from the graphs. Using the Hungarian algorithm on these probabilities we recover a mapping between the graphs. To calculate graph similarity, we combine these probabilities with edge consistency information to give a consistency measure. We analyse our approach experimentally using synthetic graphs.

This paper presents a novel visual speech recognition approach based on motion segmentation and hidden Markov models (HMM). The proposed method identifies utterances from mouth video, without evaluating voice signals. The facial movements in the video data are represented using 2D spatial-temporal templates (STT). The proposed technique combines discrete stationary wavelet transform (SWT) and Zernike moments to extract rotation invariant features from the STTs. HMMs are used as speech classifier to model English phonemes. The preliminary results demonstrate that the proposed technique is suitable for phoneme classification with a high accuracy.

Approaches based on obtaining relevant information from overwhelmingly large sets of measures have been recently adopted as an alternative to specialized features. In this work, we address the problem of finding a relevant subset of features and a suitable rotation (combined feature selection and feature extraction) as a weighted rotation. We focus our attention on two types of rotations: Weighted Principal Component Analysis and Weighted Regularized Discriminant Analysis. The objective function is the maximization of the 4 ratio. Tests were carried out on artificially generated classes, with several non-relevant features. Real data tests were also performed on segmentation of naildfold capillaroscopic images, and NIST-38 database (prototype selection).

On top of the prediction modes defined in the H.264/AVC standard, Scalable Video Coding defines prediction modes for inter-layer prediction. These inter-layer prediction modes allow the re-use of coded data from the base layer, at the cost of increasing the search space at the encoder and as a result increase the encoding time. In this paper, we investigate the relation between the coding decisions taken in the base layer and the enhancement layer. Our tests have shown that a number of relations can be clearly identified. We have observed that the co-located macroblock of a base layer macroblock coded in mode has a 40 % chance of being coded in mode as well. Further, we have observed that the mode is only used when the quantization parameter in the enhancement layer is high. For macroblocks coded in mode, the co-located macroblock in the enhancement layer will be coded in the mode when it is highly quantized (probability of 63 % to 92 % for quantization parameter 30). These observations can be used to construct a model for fast mode decision in SVC.

The use of traditional moment invariants is limited to a certain set of simple geometric transforms, such as rotation, scaling and affine transform. This paper presents a novel concept of so-called implicit moment invariants, which enable us to recognize objects under a broader set of geometric deformations.

In the present study, a new gridding method based on continuous wavelet transform (CWT) was performed. Line profiles of x and y axis were calculated, resulting to 2 different signals. These signals were independently processed by means of CWT at 15 different levels, using daubechies 4 mother wavelet. A summation, point by point, was performed on the processed signals, in order to suppress noise and enhance spot’s differences. Additionally, a wavelet based hard thresholding filter was applied to each signal for the task of alleviating the noise of the signals. 14 real microarray images were used in order to visually assess the performance of our gridding method. Each microarray image contained 4 sub-arrays, each sub-array 40x40 spots, thus, 6400 spots totally. Moreover, these images contained contamination areas. According to our results, the accuracy of our algorithm was 98% in all 14 images and in all spots. Additionally, processing time was less than 3 sec on a 1024×1024×16 microarray image, rendering the method a promising technique for an efficient and fully automatic gridding processing.

cDNA micro arrays are more and more frequently used in molecular biology as they can give insight into the relation of an organism’s metabolism and its genome. The process of imaging a micro array sample can introduce a great deal of noise and bias into the data with higher variance than the original signal which may swamp the useful information. As imperfections and fabrication artifacts often impair our ability to measure accurately the quantities of interest in micro array images, image processing for analysis of these images is an important and challenging problem. How to eliminate the effect of the noise imposes a challenging problem in micro array analysis. In this paper we implemented a novel algorithm for image sifting which could remove objects with definite size from macro array images. We used regular moving grids to sift noise object and obtained clean images for segmentation. The results have been compared with SWT, DWT and wiener filter denoising.

Terahertz Computed Tomography (THz-CT) is a form of optical coherent tomography, which offers a promising approach for achieving non-invasive inspection of solid materials, with potentially numerous applications in industrial manufacturing and biomedical engineering. With traditional CT techniques such as X-ray tomography, full exposure data are needed to produce cross sectional images, even if the region of interest is small. For time-domain terahertz measurements, the requirement for full exposure data is impractical due to the slow measurement process. In this paper, we apply a wavelet-based algorithm to locally reconstruct THz-CT images with a significant reduction in the required measurements. The algorithm recovers an approximation of the region of interest from terahertz measurements within its vicinity, and thus improves the feasibility of using terahertz imaging to detect defects in solid materials and diagnose disease states for clinical practice.