Catálogo de publicaciones - libros

In this paper, we introduce a new multiscale representation for 2-D images named the . The ICP is a decimated pyramid of complex values based on the Dual-Tree Complex Wavelet Transform (DT-CWT). The complex phases of its coefficients correspond to the angles of dominant directional features in their support regions. As a sparse representation of this information, the ICP is relatively simple to calculate and is a computationally efficient representation for subsequent analysis in computer vision activities or large data set analysis. Examples of ICP decomposition show its ability to provide an intuitive representation of multiscale features (such as edges and ridges). Its potential uses are then discussed.

The paper addresses the computer vision aspects of aligning a hydraulic cylinder prior to being hooked on a conveyer by a robotic arm. The robotic arm is programmed to assume the cylinder’s clevis hole is perpendicular to the horizontal base of the stamping station; if the cylinder is not in this orientation, the arm will unsuccessfully attempt to hook the cylinder on the conveyor line, dropping it to the concrete floor. The approach is based on the use of the Hough transform for circle detection. A camera is mounted in a rotational orientation cradle and the different camera positions result in images in which the hole is seen as an ellipse that evolves to a circle as the correct angle is reached. The paper then discusses the effect of implementing circle detection on ellipses, and takes advantage of the count in the Hough parameter space that indicates the correct position. The approach has shown to be very efficient under the restrictions of positioning the cylinder in less than 35 seconds as well as achieving orientation errors less than +/- 5°.

This paper describes a model-based method for detecting lip region from image sequences. Our approach is by Sampled Active Contour Model (S-ACM). The original S-ACM has the problem which can’t expand. To overcome this problem, we propose the elastic S-ACM. Moreover, based on the extracted lip contour, the effective delta radius features are fed to the word HMM. We recorded ten words that uses for the wheelchair control, and obtained a recognition rate of 89% with twelve features.

Real time recognition of visual face appearances (visemes) which correspond to phonemes and their speech contexts is presented. We distinguish six major classes of visemes. Features are extracted in the form of normalized image texture. The normalization procedure uses barycentric coordinates in a mesh of triangles superimposed onto a reference facial image. The mesh itself is defined using a subset of FAP points conforming with MPEG-4 standard. The elaborated classifiers were designed by PCA subspace and LDA methods. It appears that the LDA classifier outperforms subspace technique. It is better than the best subspace PCA – in recognition rate by more than 13% times (97% versus 84%) and it is more than 10 times faster (0.5 versus 7) and its time is neglected w.r.t. mouth image normalization time (0.5 versus 5).

Orthogonal moments are recognized as useful tools for object representation and image analysis. It has been shown that the recently developed discrete orthogonal moments have better performance than the conventional continuous orthogonal moments. In this paper, a new set of discrete orthogonal polynomials, namely Hahn polynomials, are introduced. The related Hahn moment functions defined on this orthogonal basis set are investigated and applied to image reconstruction. In experiments, the Hahn moments are compared with the other two discrete orthogonal moments: Chebyshev and Krawtchouk moments. The simulation results show that the Hahn moment-based reconstruction method is superior to the other two discrete orthogonal moment-based methods.

Recently semantic classification of images is of great interest for image indexing applications. On the one hand, researchers in the field of content-based image retrieval are interested in object(s) of interest in an image, which is useful for representing the image. In this paper, we present a semantic classification method of the object(s) of interest into artificial/natural classes. We first show that dominant orientation features in Gabor filtering results of artificial objects are very useful for discriminating them from natural objects. Dominant orientations in artificial object images are not confined to horizontal and/or vertical directions, while those in artificial scene images tend to be greatly confined to them. Two classification measures are proposed; the sum of sector power differences in Fourier power spectrum and the energy of edge direction histogram. They show classification accuracy of 85.8% and 84.8% on a test with 2,600 object images, respectively.

This paper proposes a novel method for the recognition of passports based on the fuzzy binarization and the fuzzy RBF network. First, for the extraction of individual codes for recognizing, this paper targets code sequence blocks including individual codes by applying Sobel masking, horizontal smearing and a contour tracking algorithm on the passport image. Then the proposed method binarizes the extracted blocks using fuzzy binarization based on the trapezoid type membership function. Then, as the last step, individual codes are recovered and extracted from the binarized areas by applying CDM masking and vertical smearing. This paper also proposes an enhanced fuzzy RBF network that adapts the enhanced fuzzy ART network for the middle layer. This network is applied to the recognition of individual codes. The results of the experiments for performance evaluation on the real passport images showed that the proposed method has the better performance compared with other approaches.

In analyzing the morphological information of objects in images, isosurfaces play important and application-independent roles. For continuous scalar field, Contour Trees have been used as a tool to select and visualize isosurfaces. However, the tree structure of contour trees is based on the critical points which does not exist in digital images. In this paper, we propose a tree structure of isosurfaces in digital images named Region-based Contour Tree. The proposed method describes a finite number of isosurfaces in digital images completely, without redundancy.

Concavity trees have been known for quite some time as structural descriptors of 2-D shape; however, they haven’t been explored further until recently. This paper shows how 2-D shapes can be concisely, but reversibly, represented during concavity tree extraction. The representation can be exact, or approximate to a pre-set degree. This is equivalent to a lossless, or lossy compression of the image containing the shape. This paper details the proposed technique and reports near-lossless compression ratios that are 150% better than the JBIG standard on a test set of binary silhouette images.

A key issue of content-based image retrieval is exploring how to bridge the gap between the high-level semantics of an image and its lower-level properties, such as color, texture and edge. In this paper, we present a new method using perceptual edge features, called generic edge tokens (GET), as image shape content descriptors for CBIR. GETs represent basic types of perceptually distinguishable edge segments including both linear and nonlinear features, which are modeled as qualitative shape descriptors based on perceptual organization principles. In the method, an image is first transformed into GET map on the fly. The base GETs can be grouped into higher-level perceptual shape structures (PSS) as additional shape descriptors. Image content is represented statistically by perceptual feature histograms (PFHs) of GETs and PSSs. Similarity is evaluated by comparing the differences between the corresponding PFHs from two images. Experimental results are provided to demonstrate the potential of the proposed method.