Papers by Keyword: Image Reconstruction

Abstract: This paper deals with a reconstruction of random media via multi-objective optimization. Two statistical descriptors, namely a two-point probability function and a two-point lineal path function, are repetitively evaluated for the original medium and the reconstructed image to appreciate the improvement in the optimization progress. Because of doubts of the weights setting in the weighted-sum method, purely multi-objective optimization routine Non-dominated Sorting Genetic Algorithm~II is utilized. Three operators are compared for creating new offspring populations that satisfy a prescribed volume fraction constraint. The main contribution is in the testing of the proposed methodology on several benchmark images.

Abstract: The article considers the reconstruction of images by Synthetic Aperture Focusing Technique (SAFT). The work compares additive and multiplicative methods for processing signals from antenna array. We have proven that the multiplicative method gives better resolution. The study includes the estimation of beam trajectories for antenna arrays using analytical and numerical methods. We have shown that the analytical estimation method allows decreasing the image reconstruction time in the case of linear antenna array implementation.

Abstract: Expectation Maximization Algorithm and the Exact Inversion Formula are two mathematical methods that have been developed for various computational applications, such as in medical imaging, nuclear industries, econometric and sociological studies, as well as chemical engineering industries. These image reconstruction methods are usually used to create the SPECT scan images. However, most of the improvement and development of the images are made by using a medical phantom, such as the human brain phantom. Here, in this paper the reconstruction of images by both algorithms are made by using a numerical phantom of laboratory scale bubble columns due to its wide application in the chemical reaction engineering studies. The results for both algorithms are compared, evaluated and discussed.

Abstract: Due to the artifact of classical image reconstruction algorithms in texture regions, this paper proposes an adaptive region demosaicking using Multi-Gradients. It first recovers green components according to the operators and the texture regions. Then the chrominance channels are reconstructed based on the interpolated green components and constant-hue rule. Experiments prove that this algorithm can improve the effectiveness from the subjective and objective.

Abstract: With the increasing development of modern social information degree, the multimedia technology and network technology have rapid development, and the society puts forward new requirements for the education industry. Accordingly, this paper uses virtual design of Flash3D animation to development and design demonstration system of track and field action. The system uses the built-in method to extract the feature of track and field action, and uses the wavelet reconstruction algorithm to reconstruct the action, so as to achieve the function of action learning. We use Adobe Flash CS65 of the system to extract 5 groups of 1025 movements and reconstruct 10 groups of 2256 movements. Comparing the reconstructed image with the actual image frames, we found the new images reconstructed by the demonstration system are consisting with the real images frame. It provides a new computer method for teaching track and field action.

Abstract: In this paper, a new face recognition method based on eigenface image reconstruction and Fisherface is proposed, it is mainly used to reduce the loss of personal characteristics. First, we can obtain the feature subspace of all the classes in training set by using the inner-classes covariance matrix as generating matrix, and so we get the eigenfaces of each person (class). Next, we use the principal component of testing set, which is obtained by mapping testing set to the feature subspace, to reconstruct the testing images. Finally, we substitute the reconstructed testing images for the original ones, and then get the recognition work completed by using Fisherface method. The simulation illustrates the effectivity of the method on the ORL face database.

Abstract: The LSQR algorithm is always used to solve the inverse problem of electrical impedance tomography (EIT). However, it always has relatively low reconstruction speed. In this paper, WALSQR (wavelet multi-resolution based Least Square QR-factorization) algorithm is proposed for EIT imaging. With the aid of wavelet transformation, the LSQR solution is obtained in the low-dimension scale space, where important information on the reconstructed image is contained. Hence the computational complexity of reconstruction is reduced without affecting the image quality. In order to verify the effectiveness of the new method, experiments of 2D and 3D EIT imaging are conducted. It lays the foundation for the study of 3D dynamic EIT image reconstruction algorithm.

Abstract: 3D animation technology is introduced into a virtual fashion show to design a new fashion show of show creation system in this paper. This system combines image processing and reconstruction technology and Matlab software, the two methods and tools. Through the change of frame vector data, it realizes the reconstruction of image. In order to improve the speed of image capture and reconstruction, PCA algorithm is adopted to reduce the dimensions of the vector data of each frame image. This system makes The three dimensional animation technology and performing choreography get together, which has achieved better artistic effect of show design, and has provided technical reference for the computer design of virtual fashion show in universities.

Abstract: In classic radial sampling scheme, angle intervals between adjacent straight sampling lines are the same. And so the coherence between the sampling matrix and the sparse matrix is high. Many pseudo-shadows will occur on the reconstructed image. For this problem, we put forward a novel sampling scheme in which the scan lines are placed in the manner of golden section angle interval. In the resulting sampling matrix, the scan lines scatter randomly on the whole. Experimental results demonstrate that the algorithm proposed in this paper has lower correlation coefficient between sampling matrix and sparse matrix than classic radial sampling scheme and random radial sampling scheme. Image can be accurately reconstructed by collecting less data than the other two schemes.

Abstract: Radioactive Particle Tracking (RPT) technique has emerged as a potential and versatile technique, both in terms of richness of information and applicability to a variety of multiphase flow systems. RPT is not an off-the-shelf technique and thus has to be developed by the intended user. This paper is intended to present a simple method and procedure for preparing suitable radioactive particles tracer (Au-198 and Sc-46) irradiated simultaneously with neutrons using TRIGA Mark II research reactor. These present work focuses on the performance evaluation of encapsulated gold and scandium particle to be used as single radioactive particle tracer using qualitative and quantitative neutron activation analysis (NAA) and X-ray micro computed tomography (micro-CT) scanner installed at Malaysian Nuclear Agency.