Catálogo de publicaciones - libros

This paper proposes a new fitting algorithm which we call Stereo Active Appearance Model (STAAM). This algorithm fits a 2D+3D Active Appearance Model to stereo images acquired from calibrated vision system and computes the 3D shape and rigid motion parameters. The use of calibration information reduces the number of model parameters, restricts the degree of freedom in the model parameters, and increases the accuracy and speed of fitting. Moreover, the STAAM uses a modified inverse compositional simultaneous update fitting algorithm to reduce the fitting computation greatly. Experimental results show that (1) the modified inverse compositional simultaneous update algorithm accelerates the AAM fitting speed while keeping its fitting accuracy, (2) the STAAM improves fitting stability using calibration information.

The problem of recognition of a polyline as a sequence of geometric primitives is important for the resolution of applied tasks such as post-processing of lines obtained as a result of vectorization; polygonal line compression; recognition of characteristic features; noise filtering; and text, symbol, and shape recognition. Here, a method is proposed for the approximation of polylines with straight segments, circular arcs, and free curves.

Low-dose X-ray image sequences exhibit severe signal-dependent noise that must be reduced in real-time while, at the same time, preserving diagnostic structures and avoiding artifacts. We propose three different methods with applications beyond medical image processing. Major contributions are innovative motion detection based on independent binarization of positive and negative temporal differences, real-time multiscale nonlinear diffusion in the presence of severe signal-dependent noise, and multi-resolution inter-scale correlation in shift-dependent pyramids. All methods exhibit excellent performance over a broad range of noise, detail, and contrast levels. As performance in medical imaging depends to a large degree on the type of intervention and individual preferences of medical staff, no method is generally superior and all methods are considered for the next generation of fluoroscopy systems.

A widely used practice in Burn Wounds Centres is the incorporation of a photograph of the burned area into the patient’s clinical history. This photograph is used as a reference for each revision of the diagnosis or the therapeutic plan. This article presents the results of the evaluation of various fuzzy clustering algorithms applied to the segmentation of burn wounds images. The study compares recent and classical algorithms in order to establish a better comparison between the benefits of more complex techniques for pixel classification. Our final purpose is to develop a module that provides the medical expert with information on the extension of the burned area.

In this paper a new approach to the subject of burn image classification is presented. When the physicians determine that a certain burn is superficial or deep, they are, unconsciously, examining some features of the burn difficult to translate into a physical measure by them. To implement a CAD tool, we need to identify these features in order to perform the same operation in an automatic way. To this aim, we have designed an experiment following the Recommendation ITU-R BT. 500-10, in which we ask 8 experts in burn images to assess the similarities of a group of selected images. Afterwards a mathematical analysis, based in the multidimensional scaling (MDS), has let us identify the numerical features that have lead experts to this classification. This suggests a promising way to automatically classify burn images.

The paper presents the methodology developed to solve the class imbalanced problem that occurs in the classification of Thin-Layer Chromatography (TLC) images. The proposed methodology is based on re-sampling, and consists in the undersampling of the majority class (normal class), while the minority classes, which contain Lysosomal Storage Disorders (LSD) samples, are oversampled with the generation of synthetic samples. For image classification two approaches are presented, one based on a hierarchical classifier and another uses a multiclassifier system, where both classifiers are trained and tested using balanced data sets. The results demonstrate a better performance of the multiclassifier system using the balanced sets.

3D ultrasound is a unique medical imaging modality for observing the growth and malformation of the fetus. But it is necessary to enhance its visual quality by filtering to reduce speckle noise and artifacts. Because imaging of fetuses takes place real time, these processes must also be fast. Previous methods have limited speed, quality, or are only applicable to 2D. We propose a new 3D filtering technique for 3D US fetus volume data which classifies the volume according to local coherence and applies different filters to the volume of interest and to the rest of the 3D image. The volume of interest, which contains the fetus, is determined automatically from key frames, and is processed using a nonlinear coherence enhancing diffusion (NCED) filter. Our method enhances 3D US fetus images more effectively than previous techniques, runs more quickly, and reduces the number of artifacts because it is a true extension to 3D.

A computational framework based on particle filter is proposed for fully automatic determination of morphological parameters of proximal femur from calibrated fluoroscopic images. In this framework, the proximal femur is decomposed into three components: (1) femoral head, (2) femoral neck, and (3) femoral shaft, among which structural constraints are defined according to the anatomical structure of the proximal femur. Each component is represented by a set of parameters describing its three-dimensional (3D) spatial position as well as its 3D geometrical shape. The constraints between different components are modeled by a rational network. Particle filter based inference is then used to estimate those parameters from the acquired fluoroscopic images. We report the quantitative and qualitative evaluation results on 10 dry cadaveric femurs, which indicate the validity of the present approach.

Gastric cancer is a great part of the cancer occurrence and the mortality from cancer in Korea, and the early detection of gastric cancer is very important in the treatment and convalescence. This paper, for the early detection of gastric cancer, proposes the analysis system of an endoscopic image of the stomach, which detects the abnormal region by using the change of color in the image and by providing the surface tissue information to the detector. While advanced inflammation and cancer may be easily detected, early inflammation and cancer are difficult to detect and requires more attention to be detected. This paper, at first, converts the endoscopic image to the image of the IHb(Index of Hemoglobin) model and removes noises incurred by illumination and, automatically detects the regions suspected as cancer and provides the related information to the detector, or provides the surface tissue information for the regions appointed by the detector. This paper does not intend to provide the final diagnosis of the detected abnormal regions as gastric cancer, but it intends to provide a supplementary mean to reduce the load and mistaken diagnosis of the detector, by automatically detecting the abnormal regions not easily detected by the human eye and this provides additional information for the diagnosis. The experiments using practical endoscopic images for performance evaluation showed that the proposed system is effective in the analysis of endoscopic image of the stomach.

Emission tomography images are degraded due to the presence of noise and several physical factors, like attenuation and scattering. To remove the attenuation effect from the emission tomography reconstruction, attenuation correction factors (ACFs) are used. These ACFs are obtained from a transmission scan and it is well known that they are homogeneous within each tissue and present abrupt variations in the transition between tissues. In this paper we propose the use of compound Gauss Markov random fields (CGMRF) as prior distributions to model homogeneity within tissues and high variations between regions. In order to find the maximum a posteriori (MAP) estimate of the reconstructed image we propose a new iterative method, which is stochastic for the line process and deterministic for the reconstruction. We apply the ordered subsets (OS) principle to accelerate the image reconstruction. The proposed method is tested and compared with other reconstruction methods.