3:30, IMDSP-P9.1
EXTRACTING PERSONAL CHARACTERISTICS FROM HUMAN MOVEMENT
J. HOSHINO
In this paper, we propose a new method for extracting personal characteristics from 3D body movement. We introduce the eigen action space to represent the personal characteristics. First, we estimate the average action from a set of 3D pose parameters from different people. Then we created the eigen action space from the covariance matrics of 3D pose parameters using KL transform. Because the eigen action space consists of orthogonal base vectors, 3D pose parameters of a person is represented as a point. Similarity measure is calculated from points in action eigen space. Also, actions with new personal characteristics can be reconstructed by sampling new points in the eigen action space.

3:30, IMDSP-P9.2
ANALYSIS OF ECHOES IN SINGLE-IMAGE RANDOM-DOT-STEREOGRAMS
M. LAU, C. KWONG
Three-dimensional depth information of a surface can be encoded in a two-dimensional image called single-image random-dot-stereograms or, more widely known, autostereograms. It is achieved by using the correlations of pixels in the horizontal direction. Using the correspondences between pixels in human brains or computer algorithms, surfaces can be reconstructed from autostereograms. However, in some cases, the reconstructed surfaces are not unique because of ``echoes''. In the presence of echoes, reconstruction of the original surface from an autostereogram cannot be guaranteed since no cue of the original surface is available in autostereograms. In this paper, the causes of echoes are investigated and conditions for echo-free reconstructions are derived. Based on these conditions, an improved autostereogram generation algorithm is proposed to guarantee echo-free autostereograms. Besides, the surface reconstruction algorithm is modified such that the originally encoded surfaces can always be reconstructed from echo-free autostereograms.

3:30, IMDSP-P9.3
WEIGHTED LEAST SQUARES METHOD FOR THE APPROXIMATION OF DIRECTIONAL DERIVATIVES
M. TICO, P. KUOSMANEN
Using the facet model we design a family of filters for the approximation of partial derivatives of the digital image surface. Prior information (e.g., local dominant orientation) are incorporated in a two dimensional weight function. A weighted least squares estimation of the facet parameters is applied in order to design the proposed filters. Exemplary application of the proposed filters to fingerprint image segmentation is also presented.

3:30, IMDSP-P9.4
NEW DISPARITY MAP ESTIMATION USING HIGHER ORDER STATISTICS
M. RZIZA, D. ABOUTAJDINE, L. MORIN, A. TAMTAOUI
This paper presents a new algorithm of disparity map estimation. Originality of this method lies in the process of dense disparity map estimation using the dynamic programming constrained by interest points and using Higher Order Statistics (HOS) criteria for matching noisy images. Experiments with noisy real images have validated our method and have clearly shown the improvement over the existing ones. The dense disparity map obtained is more reliable when compared to the similar Second-Order Statistics (SOS) based dynamic programming and HOS based correlation methods.

3:30, IMDSP-P9.5
GEOMETRIC ALIGNMENT OF TWO OVERLAPPING RANGE IMAGES
M. RODRIGUES, Y. LIU, Q. WEI
In this paper we propose a novel geometric method for the alignment of two overlapping range images. The method first employs the traditional ICP criterion to establish a set of possible correspondences and then refine these correspondences using geometric constraints derived from properties of reflected correspondence vectors. In this way, the method overcomes a major limitation of the ICP criterion which is the introduction of false matches in almost every iteration of the alignment. For an accurate estimation of the geometric parameters of interest, the Monte Carlo method is used in conjunction with a median filter. Finally, the quaternion method is used to estimate the motion parameters based on the refined set of correspondences. Experimental results based on both synthetic data and real images show that the proposed method can effectively align two overlapping range images with a small motion.

3:30, IMDSP-P9.6
IMAGE REGISTRATION BASED ON BOTH FEATURE AND INTENSITY MATCHING
Y. JIANCHAO
Image registration is one of the most important taks in image processing. In this paper, a new algorithm for image registration was developed based on both feature and intensity matching. The algorithm utilises a parametric projective model accounting for geometrical variation and a polynomial model with a small number of polynomial coefficients explicating the smooth spatially varying illumination variation. The initial projective model parameters are first estimated by using feature-based approach. Subsequently, the coefficients of the illumination model are determined simultaneously with final projective transformation parameters through the process of intensity matching. The experimental results demonstrated the algorithm is of robustness, effciency and accuracy.

3:30, IMDSP-P9.7
CONTENT-BASED REPRESENTATION OF COLOUR IMAGE SEQUENCES
I. KOMPATSIARIS, M. STRINTZIS
In this paper, a procedure is described for the spatiotemporal segmentation and tracking of objects in colour image sequences. For this purpose, we propose the novel procedure of K-Means with connectivity constraint algorithm as a general segmentation algorithm combining several types of information including colour, motion and compactness. A new colour distance is also defined for this algorithm. The regularisation parameters are evaluated automatically using the min-max criterion. In this algorithm, the use of spatiotemporal regions is introduced since a number of frames is analyzed simultaneously and as a result the same region is present in consequent frames. Experimental results on real and synthetic colour data demonstrate the performance of the data.

3:30, IMDSP-P9.8
MULTI-LINE FITTING USING POLYNOMIAL PHASE TRANSFORMS AND DOWNSAMPLING
K. ABED-MERAIM, A. BEGHDADI
A new signal processing method is developed for solving the multi-line fitting problem in a two dimensional image. We first reformulate the former problem in a special parameter estimation framework such that a first order or a second order polynomial phase signal structure is obtained. Then, the recently developed algorithms in that formalism (and particularly the downsampling technique for high resolution frequency estimation) can be exploited to produce accurate estimates for line parameters. This method is able to estimate the parameters of parallel lines with different offsets and handles the quantization noise effect which can not be done by the sensor array processing technique introduced by Aghajan et al. Simulation results are presented to demonstrate the usefulness of the proposed method.