IMDSP-5.1

MULTIFRAME INTEGRATION VIA THE PROJECTIVE TRANSFORMATION WITH AUTOMATED BLOCK MATCHING FEATURE POINT SELECTION
Richard R Schultz (University of North Dakota), Mark G Alford (Air Force Research Laboratory/IFEA)

A subpixel-resolution image registration algorithm based on the nonlinear projective transformation model is proposed to account for camera translation, rotation, zoom, pan, and tilt. Typically, parameter estimation techniques for transformation models require the user to manually select feature points between the images undergoing registration. In this research, block matching is used to automatically select correlated feature point pairs between two images, and these features are used to calculate an iterative least squares solution for the projective transformation parameters. Since block matching is capable of estimating accurate translation motion vectors only in discontinuous edge regions, inaccurate feature point pairs are statistically eliminated prior to computing the least squares parameter estimate. Convergence of the projective transformation model estimation algorithm is generally achieved in several iterations. After subpixel-resolution image registration, a high-resolution video still may be computed by integrating the registered pixels from a short sequence of low-resolution image sequence frames.

IMDSP-5.2

THREE-DIMENSIONAL SHAPE RECOVERY FROM FOCUSED IMAGE SURFACE
Tae S. Choi, Muhammad Asif, Joungil Yun (Dept. of Mechatronics, Kwangju Institute of Science and Technology)

A new method for the three-dimensional shape recovery from image focus is proposed. The method is based on approximation of the Focus Image Surface(FIS) by a piecewise curved surface which tracks the realistic FIS in image space. The piecewise curved surface is estimated by interpolation using the Lagrangian polynomial. The new method has been implemented on a prototype camera system. The experiments and their results are provided and discussed. The experimental results show that the new method gives more accurate results than the previous methods.

IMDSP-5.3

Minimization of Weighted Sensitivity for 2-D State-Space Digital Filters Described by the Fornasini-Marchesini Second Model
Takao Hinamoto, Akimitsu Doi, Shuichi Yokoyama (Faculty of Engineering, Hiroshima University)

This paper considers the problem of minimizing the weighted coefficient sensitivity for 2-D state-space digital filters described by the Fornasini-Marchesini (F-M) second model. First, a simple technique is presented for obtaining a set of filter structure with very low weighted L1/L2-sensitivity. Next, an iterative procedure is applied to obtain the optimal coordinate transformation that minimizes the weighted L2-sensitivity measure. This is based on the matrix Riccati differential equation. Finally, a numerical example is given to illustrate the utility of the proposed technique.

IMDSP-5.4

A New View of Shape
Vinod Chandran, Wageeh W Boles (Queensland University of Technology, Brisbane, Australia)

A new theory that qualitatively and quantitatively describes shape is presented. Signals are treated as ordered sets and shape as a property of the order. Measures of shape are derived starting from simple two element sets. It is shown that the degree of asymmetry or relative contrast at various scales is a sufficient descriptor of shape. The shape of a signal is a composite of the shape of all possible ordered subsets. A distinction is drawn between statistical and shape measures within a unified theoretical framework which may make it possible to compare diverse pattern recognition approaches in terms of robustness and discrimination power.

IMDSP-5.5

COMPARISON OF SECOND AND THIRD ORDER STATISTICS BASED ADAPTIVE FILTERS FOR TEXTURE CHARACTERIZATION
Mounir Sayadi (ESSTT, University of Tunis, Tunisia), Mohamed Najim (Equipe Signal & Image and GDR-ISIS-CNRS, ENSERB, France)

In the framework of parametric texture modeling, a question arises: are adaptive approaches based on higher order statistics (HOS) more appropriate to characterize texture models than those based on second order statistics (SOS)? In order to give some responses to this question, we have compared two fast adaptive filters for texture characterization: the 2-D FLRLS filter (2-D Fast Lattice Recursive Least Square) based on SOS only and the 2-D OLRIV filter (2-D Overdetermined Lattice Recursive Instrumental Variable) based on third order statistics. Extensive experiments to study the characterization performance of each filter are presented and interpreted. They show that the 2-D FLRLS filter provides a very good performance for texture characterization, even when with important noise. Furthermore, the third order based algorithm presents higher variance than second order one. We believe that for 2-D adaptive modeling, there is no advantage to use a HOS based adaptive algorithm for characterizing textures.

IMDSP-5.6

Object Rotation Axis from Shading
J�rgen Stauder (IRISA/INRIA)

In this paper, a non-complex estimator is developed for the tilt angle of the rotation axis of an object that is illuminated by a point light source. The tilt angle defines the orientation of the 2D projection of the 3D object rotation axis in the image plane and is a strong clue for image understanding. The estimator evaluates two images of a video image sequence showing a moving object. Additionally, only a displacement vector field and the 2D object silhouette are required. No 3D information is required. Therefore, the object is assumed to be rigid, to be matte, to have equally distributed surface normals and to be illuminated by a distant point light source and ambient light. For estimation, the displaced frame ratio (DFR), i.e. the frame ratio after motion compensation, is evaluated statistically. The DFR depends only on the photometric effect of temporally changing object shading. Experimental results with real images show the proper performance of the derived estimator for real objects. A demo is in http://www.irisa.fr/prive/Jurgen.Stauder

IMDSP-5.7

Texture analysis of an image by using a rotation-invariant model
Christophe Rosenberger, Kacem Chehdi, Claude Cariou, Jean-marc Ogier (ENSSAT - LASTI)

Texture analysis is an important problem in image processing because it conditions the quality of image segmentation and interpretation. We propose in this communication a texture model which is invariant by rotation and whose parameters allow to characterize at the same time the type of texture and its tonal primitive. The originality of the model proposed lies in the use of the Wold decomposition to modelize the 1D normalized autocovariance. This function is computed from the 2D normalized autocovariance of a texture. Finally, parameters of the model are estimated by using a genetic algorithm. Experimental results on textures from the Brodatz album and synthetic textures show a modeling error lower than 0.06.

IMDSP-5.8

Fast and Exact Signed Euclidean Distance Transformation with Linear Complexity
Olivier Cuisenaire, Benoit Macq (Communications and Remote Sensing Laboratory, Universit� catholique de Louvain, Belgium)

We propose a new signed or unsigned Euclidean distance transformation algorithm, based on the local corrections of the well-known 4SED algorithm of Danielsson. Those corrections are only applied to a small neighborhood of a small subset of pixels from the image, which keeps the cost of the operation low. In contrast with all fast algorithms previously published, our algorithm produces perfect Euclidean distance maps in a time linearly proportional to the number of pixels in the image. The computational cost is close to the cost of the 4SSED approximation.

IMDSP-5.9

Scale Detection Based on Statistical Characteristics of Edges in the Scale Space
Nawapak Eua-Anant, Lalita Udpa (Department of Electrical and Computer Engineering, Iowa State University, Ames, IA 50011)

A study of the behavior of edges in the scale space is presented in this paper. Three statistical parameters, intensity average, root-mean-square, and variance of pixel intensity in edge images are used as the basis for evaluating visual quality of edge images. A parametric family of edge images containing edge information at several scales is used to create the edge statistic (ES) curves. It is found that the number of local minima in ES curves is related to the number of scales in an image. This leads to an automated method for detecting scales in an image and a criteria for selecting optimal parameters for a multiscale edge operator.

IMDSP-5.10

The Beltrami geometrical framework for color image processing
Nir Sochen, Yehoshua Y Zeevi (Faculty of Electrical Engineering, Technion-Israel Institute of Technology, Israel)

The Beltrami geometrical framework for scale-space flows is generalized to non-trivial color space geometries and implemented in analysis and processing of color images. We demonstrate how various models of color perception, interpreted as geometries of the color space, result in different enhanced processing schemes.

< IMDSP-4 IMDSP-6 >

Last Update: February 4, 1999 Ingo Höntsch