Catálogo de publicaciones - libros

This paper introduces a new concept within shadow segmentation for usage in shadow removal and augmentation through construction of an alpha overlay shadow model. Previously, an image was considered to consist of shadow and non-shadow regions. We construct a model that accounts for sunlit, umbra and penumbra regions. The model is based on theories about color constancy, daylight, and the geometry that causes penumbra. The behavior of the model is analyzed and a graph cut energy minimization is applied to estimate the alpha parameter. The approach is demonstrated on natural complex image situations. The results are convincing, but the alpha gradient in penumbra must be improved.

In this paper we propose an algorithm for the sub-pixel edge detection using a B-spline model. In contrast to the usual methods which are generally sensitive to local perturbations, our approach is based on a global computation of the edge using a Maximum Likelihood rule. In the proposed algorithm the likelihood of the observations is explicitly computed, it ensures the filtering of the noisiest data. Experiments are given and show the adequacy and effectiveness of this algorithm.

In this paper we describe a method for re-mapping animation parameters between multiple types of facial model for performance driven animation. A facial performance can be analysed automatically in terms of a set of facial action trajectories using a modified appearance model with modes of variation encoding specific facial actions. These parameters can then be used to animate other appearance models, or 3D facial models. Thus, the animation parameters analysed from the video performance may be re-used to animate multiple types of facial model.

We demonstrate the effectiveness of our approach by measuring its ability to successfully extract action-parameters from performances and by displaying frames from example animations.

In this paper we propose an original method to animate a crowd of virtual beings in a virtual environment. Instead of relying on models to describe the motions of people along time, we suggest to use knowledge on the dynamic of the crowd acquired from videos of real crowd situations. In our method this information is expressed as a time-varying motion field which accounts for a continuous flow of people along time. This motion descriptor is obtained through optical flow estimation with a specific second order regularization. Obtained motion fields are then used in a classical fixed step size integration scheme that allows to animate a virtual crowd in real-time. The power of our technique is demonstrated through various examples and possible follow-ups to this work are also described.

In this paper, we present a wavelet-based progressive compression method for isomorphic 3-D mesh sequence with constant connectivity. Our method reduces the spatial and temporal redundancy by using both spatial and temporal wavelet analysis. To encode geometry information, each mesh frame is decomposed into a base mesh and its spatial wavelet coefficients of each resolution level by spatial wavelet analysis filter bank. The spatially transformed sequence is decomposed into several sub-band signals by temporal wavelet analysis filter bank. The resulting signal is encoded by using an arithmetic coder. Since an isomorphic mesh sequence has the same connectivity over all frames, the connectivity information is encoded only for the first mesh frame. The proposed method enables both progressive representation and lossless compression in a single framework by multi-resolution wavelet analysis with a perfect reconstruction filter bank. Our method is compared with several conventional techniques including our previous work.

The importance of information technology and ubiquitous computing are gained in the agricultural area. A remote monitoring tool, namely Field Servers, is developed and used in recent agricultural industries. The Field Server can handle weather data, measuring and taking photo images for analyzing visualized information. However, image data acquired by the Field Server is not adequately used for automatic control in current systems even though the conditions of plants, pests and thieves can be detected from the images. The purpose of this research is to develop an application which controls peripherals on the basis of features extracted from image data. As our first proposal we developed the farmer support system which pours water to wilting plants automatically. In this system, four indicators are used to detect the wilt of plants that have dense or sparse leaves. Two experiments were employed on using plants which are observed by a Field Server located outside. The experimental results prove that we can detect the wilt of plants by use of the proposed system.

Background subtraction methods have been used to obtain human silhouettes for gesture and gait recognition. However, background subtraction in pixel units is prone to error which decreases recognition performance significantly. In this paper we propose a novel background subtraction method that extracts foreground objects in region units. Together with the background model, an object’s color and movement information are used to obtain the effective region object likelihood. Then an adaptive region decision function determines the object regions. Also, the sequential version of Horprasert’s algorithm[2] is presented.

The fixed mean shape that is built from the statistical shape model produces an erroneous feature extraction result when ASM is applied to multi-pose faces. To remedy this problem the mean shape vector which is similar to an input face image is needed. In this paper, we propose the adaptive mean shape to extract facial features accurately for non frontal face. It indicates the mean shape vector that is the most similar to the face form of the input image. Our experimental results show that the proposed method obtains feature point positions with high accuracy and significantly improving the performance of facial feature extraction over and above that of the original ASM.

We present the design of face identification system that can run in real time environment. We use multiple contexts to optimize the face recognition performance in real time. Initially different illumination environments are modeled as context using unsupervised learning and accumulated as context knowledge. Optimization parameters for each context are learned using Genetic Algorithm (GA).GA search the optimization parameter so as to minimize the effect of illumination variation. These weight parameters are used during similarity match of face images in real time recognition. Gabor wavelet is used for facial feature representation. Experiment is done using real time face database containing images taken under various illumination conditions. The proposed context aware method has been shown to provide superior performance than the method without using context awareness.

This paper presents a computerized bone age assessment method using discrete cosine transform (DCT) and Fisher’s linear discriminant analysis (FLD or LDA). Bone age assessment using a radiograph of the left hand is a common procedure in pediatric radiology. In the proposed method, DCT and LDA are applied to the epiphyseal regions segmented from a radiograph of the left hand. The extracted LDA coefficients are compared with features stored in the database, and then the bone age of the given radiograph is estimated. In experiments on 396 radiographs of the left hand collected at Hanyang University Medical Center, the proposed method shows an average error of 0.6 years and an accuracy of 89.71%.