Wyniki wyszukiwania - Biblioteka Nauki

1

Essential representation and calibration of rigid body transformations

100%

Liu Y. , Rodrigues M. A.

Machine Graphics and Vision

|

2000

|

tom Vol. 9, No. 1/2

123-138

EN

In [20], we extended Chasles' screw motion concept to computer vision calibration tasks by investigating geometric properties of image correspondence vectors. In this paper, we analyse geometric properties of reflected correspondence vectors synthesised into a single coordinate frame leading to the development of two representation theorems of rigid transformations. The theorems and their corollaries provide a novel method to identify the original transformation and an insight into the relationships between the representation of rigid transformation, rigid constraints, and the transformation parameters. The analysis provides a framework for the development of novel calibration algorithms as demonstrated by two new proposed algoriths. For a comparative study of performance, we also implemented another well known calibration procedure based on the constraint least squares (CLS) ( [4] ) method. The experimental results have shown that the proposed algorithms are more accurate that the CLS algorithm when the image data are corrupted by both noise and outliers.

2

Range resolution improvement of range-gated vision system in backscattering hazy environments

100%

Park S.-K. , Choi I.-Y. , Baik S.-H. , Jeong K.-M.

Optica Applicata

|

2017

|

tom Vol. 47, nr 4

533--543

EN

A range-gated vision system simultaneously provides two-dimensional and range images because its light intensity contains the reflectance as well as depth information. The range-resolution of the system is usually inversely proportional to the induced backscattering noise. In this paper, a range imaging technique is proposed to precisely measure range information from highly backscattering foggy environments. A windowed center-of-mass position extracted from the peak area of a cross-correlation signal of two signals, a Gaussian window signal in reduced size and a range-gated signal according to distance, is adopted as the range depth. The proposed measuring technique provides more robust and more precise range information than conventional measuring techniques for hazy targets by virtue of the reduction of backscattering bias noise usually induced by airborne particles. The experimental results and the signal processing procedures to acquire precise range information from hazy targets are described in this paper.

3

Facial surface reconstruction in 3D format

100%

Shchegoleva N.

Journal of Theoretical and Applied Computer Science

|

2012

|

tom Vol. 6, nr 4

37-50

EN

This article presents a method of improving the quality of a reconstructed 3D surface of a face or its parts using the inverse distance method (IDM). The proposed method eliminates undesired holes in the range image, which appear due to significant omissions of measurements from the face scanner. Application of the method allows to build a 3D surface with the desired resolution, without solving the problem of triangulation therefore reducing the computational cost. The results of the experiments show the efficiency and the quality of the reconstructions obtained using the proposed algorithm.

4

Fast range image registration using distance sampling

86%

Kim S. H. , Park R. H.

Machine Graphics and Vision

|

2001

|

tom Vol. 10, No. 2

115-131

EN

This paper proposes a fast range image registration algorithm, in which control points are sampled on the basis of distance from the geometric origin of an object. The sampled point is assumed to be in the same region if the quantization error in the three-dimensional (3-D) space is less than a threshold. For range image registration, finding matching points anly in the same region reduces the computation time greatly. Experiments with various synthetic and real images show that the accuracy of registartion parameters is improved with a low computational load.

5

Scatter matrix based iterative structure and pose estimation from a single image

86%

Rodrigues M. A. , Liu Y. , Wei Q.

Machine Graphics and Vision

|

2000

|

tom Vol. 9, No. 1/2

525-538

EN

It has been demonstranted that the constraint least squeres (CLS) method yields accurate and robust motion parameter estimation from range images ([16,20]) assuming that the image data are corrupted by noise only without outliers. Also, it has been shown that scatter matrix based correspondenceless motion estimation from range images ([4]) is mathematically well defined. In this paper, we propose a new accurate and robust method for pose and strycture estimation from perspective images that draws on the good features of scatter matrix and constraint least squares methods. For reasons of parciticality and efficiency, the method first iteratively estimates the depth of some control points based on the constraints induced from the scatter matrix based correspondenceless motion estimation algorithm and then provides a closed from estimation of the depths of the other points based on rigid constraints. Finally, the method estimates pose parameters by the constraint least squares. For a comparative study of performance, we also omplemented a linear algorithm for pose estimation and a motion estimation algorithm based on epipolar geometry, [2, 24]. Experimental results based on both synthetic data and real images shown the relative superior performance of the proposed algoithm.

6

Impulsive noise cleaning by smoothing of images using fast median method (FMM)

72%

Velu C. M. , Prabakaran N. , Vivekanandan P.

Image Processing & Communications

|

2008

|

tom Vol. 13, no 3-4

19-30

EN

The objective of this paper is to propose new approaches FMM for removing impulsive noise in images so as to smooth the image for further processing. The various methods like Laplacian filter. Gaussian filter. Mean filter, Mode filter, Maximum filter, Minimum filter, Hybrid filter, Adaptive filter, Low pass filter, High pass filter, Hybrid filter and Uniform filter are discussed. In this paper, a new approach is proposed to use the FMM, which is obtained by computing median in each row, again median is calculated for medians of each row of a convolution mask. In this method, for example, for a 3 x 3 mask, left and right gray values are loaded along with center pixel of each row or each column and these values are sorted in order. The new center value is considered as median value, with this new value, smoothing of an image is achieved. In this method, a smoothed image is computed by employing just an auxiliary function swap, which yields less computation time.