Wen-Chen Huang

Adaptive-size meshes for rigid and nonrigid shape analysis and synthesis

A physically based modeling method that uses adaptive-size meshes to model surfaces of rigid and nonrigid objects is presented. The initial model uses an a priori determined mesh size. However, the mesh size increases or decreases dynamically during surface reconstruction to locate nodes near surface areas of interest (like high curvature points) and to optimize the fitting error. Further, presented with multiple 3-D data frames, the mesh size varies as the data surface undergoes nonrigid motion. This model is used to reconstruct 3-D surfaces, analyze the nonrigid motion, track the corresponding points in nonrigid motion, and create graphic animation and visualization. The method was tested on real range data, on simulated nonrigid motion, and on real data for the left ventricular motion. >

Adaptive-size physically-based models for nonrigid motion analysis

Adaptive-size physically based models suitable for nonrigid motion analysis are presented. The mesh size increases or decreases dynamically during the surface reconstruction process to locate nodes near surface areas of interests (like high curvature points) and to optimize the fitting error. A priori information about nonrigidity can be included so that the surface model deforms to fit moving data points while preserving some basic nonrigid constraints (e.g. isometry or conformality). Implementation of the proposed algorithm with and without isometric/conformal constraints is presented. Performance and accuracy of derived algorithms are demonstrated on data simulating deforming ellipsoidal and bending planar shapes. The algorithm is applied to the real range data for bending paper and to volumetric temporal left ventricular data.<<ETX>>

Recurrent nasal tumor detection by dynamic MRI

The purpose of this research is to detect and enhance the recurrent nasal tumor region by computing the relative intensity difference between consecutive MR images after using a contrast agent. In this article, we apply a relative signal increase model to recognize a recurrent nasal tumor by dynamic MR images. A robust estimation technique is used to deal with matching corresponding points among different images. The active contour technique is applied to refine automatically the region of interest and obtain a more precise definition of the area of interest. The quantitative evaluation of dynamic MR data is modeled by fitting three-parameter time-intensity curves.

Motion estimation from scaled orthographic projections without correspondences

Abstract This paper presents two new eigenstructure-based algorithms for estimating the motion parameters of a rigid object from scaled orthographic projections. The approach requires neither point correspondences between time frames nor between projections at each time. 2D statistics from projections are utilized to recover 3D statistics, which are used for motion estimation. This results in significant speed-up compared to the approach where point correspondences need to be found. The first algorithm is suitable for general point configuration and requires three scaled orthographic projections at each time frame. The second algorithm is suitable for plamar configuration of data points and requires only two orthographic projections at each time instant. Results of simulated and real image experiments demonstrate the performance and efficiency of the algorithms, as well as showing their limitations.

Nonlinear finite element methods for nonrigid motion analysis

The motion of nonrigid or deformable bodies has been studied in the field of engineering mechanics and applied mathematics for years with great success. In computer vision, the application of engineering mechanics for 3D shape fitting and motion analysis is generally called utilizing deformable shape models or physically-based modeling (usually using linear FEM). Since many real-world materials behave non-linearly when they are subjected to large deformations, one cannot expect a linear FEM to be a good model of many materials undergoing large deformation. This severely restricts the applications of FEM in many nonrigid notion analysis problems. In this paper we propose the use of the nonlinear finite element modeling for the nonrigid motion analysis

Left ventricle motion modeling and analysis by adaptive-size physically based models

This paper presents a new physically based modeling method which employs adaptive-size meshes to model left ventricle (LV) shape and track its motion during cardiac cycle. The mesh size increases or decreases dynamically during surface reconstruction process to locate nodes near surface areas of interest and to minimize the fitting error. Further, presented with multiple 3-D data frames, the mesh size varies as the LV undergoes nonrigid motion. Simulation results illustrate the performance and accuracy of the proposed algorithm. Then, the algorithm is applied to the volumetric temporal cardiac data. The LV data was acquired by the 3-D computed tomography scanner. It was provided by Dr. Eric Hoffman at University of Pennsylvania Medical school and consists of 16 volumetric (128 by 128 by 118) images taken through the heart cycle.

Sampling and surface reconstruction with adaptive-size meshes

This paper presents a new approach to sampling and surface reconstruction which uses the physically based models. We introduce adaptive-size meshes which automatically update the size of the meshes as the distance between the nodes changes. We have implemented the adaptive-size algorithm to the following three applications: (1) Sampling of the intensity data. (2) Surface reconstruction of the range data. (3) Surface reconstruction of the 3-D computed tomography left ventricle data. The LV data was acquired by the 3-D computed tomography (CT) scanner. It was provided by Dr. Eric Hoffman at University of Pennsylvania Medical school and consists of 16 volumetric (128 X 128 X 118) images taken through the heart cycle.

Nonrigid motion analysis using nonlinear finite element modeling

In this paper, we consider the problem of tracking elastic objects with known material properties. That corresponds to utilization of specific a priori knowledge in nonrigid motion analysis. We propose the utilization of nonlinear finite element methods (FEM) for point correspondence recovery. We deal with range data, i.e., assume that surface points before and after motion are available from the sensing system. Nonlinear FEM not only allows for modeling of nonlinear material properties but also for large deformation between consecutive time frames (which are main limitations of widely utilized linear FEM). We propose a new algorithm for the point correspondence recovery in nonrigid motion using assumptions of known material properties and single (but not known) force applied to the object. Simulations and experimental results demonstrate the performance of the proposed algorithm.

Analysis of Intensity and Range Image Sequences Using Adaptive-Size Meshes

Abstract This paper presents a new technique for reconstructing and analyzing sequential images using adaptive-size physically based models. In this method, the mesh size increases or decreases dynamically during reconstruction to locate nodes near surface areas of interest (such as high curvature points) and to optimize the fitting error. Image sequences of range and intensity data are used to demonstrate the power of the technique for shape estimation. Since the consecutive frames are often similar, the sequential image can be efficiently reconstructed by the adaptive-size meshes. In addition, a priori information about nonrigidity can be included so that surface model deforms to fit moving data points while preserving some basic nonrigid constraints (e.g., isometry or conformality). Several intensity and range image sequences are reconstructed efficiently using the adaptive-size meshes. The accuracy of the model is estimated in order to demonstrate the performance of this algorithm. Implementation of the proposed algorithm with and without isometric/conformal constraints is presented. The tracking of corresponding nodes using adaptive-size meshes on the face image sequences is also presented. Performance and accuracy of derived algorithms are demonstrated on simulated data of deforming ellipsoidal and bending planar shapes. Then the algorithm is applied to real range data for bending paper and to volumetric temporal left ventricular data.

Integration of multiple levels of contour information for Chinese-character stroke extraction

In this paper, we present a new stroke extraction algorithm that integrates all levels of contour information including boundary points, dominant points, corner points, segments, cross section sequence graph, character structure to extract strokes of Chinese characters. In the algorithm, first, the boundary points are extracted, then the dominant and corner points are detected. Third, the character structure including singular and regular regions are extracted by the contour information and a modified cross section sequence graph (CSSG). Finally, a Bezier curve taking dominant points and corner points as inputs is used to check the continuity of strokes. Experimental results show that the proposed algorithm can correctly extract the strokes up to 95% from a printed and handwritten test samples based on the human perception. This research provides a solid basis for the structural matching of Chinese characters.