• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.4
• /
• pp.1110-1127
• /
• 2022

To improve the training efficiency and generalization performance of a support vector machine (SVM) in a large-scale set, an optimal SVM learning method based on adaptive sparse sampling and the granularity shift factor is presented. The proposed method combines sampling optimization with learner optimization. First, an adaptive sparse sampling method based on the potential function density clustering is designed to adaptively obtain sparse sampling samples, which can achieve a reduction in the training sample set and effectively approximate the spatial structure distribution of the original sample set. A granularity shift factor method is then constructed to optimize the SVM decision hyperplane, which fully considers the neighborhood information of each granularity region in the sparse sampling set. Experiments on an artificial dataset and three benchmark datasets show that the proposed method can achieve a relatively higher training efficiency, as well as ensure a good generalization performance of the learner. Finally, the effectiveness of the proposed method is verified.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.10
• /
• pp.5015-5038
• /
• 2018

In this paper, a novel spectral-spatial joint sparse representation algorithm for hyperspectral image classification is proposed based on multi-layer superpixels in various scales. Superpixels of various scales can provide complete yet redundant correlated information of the class attribute for test pixels. Therefore, we design a joint sparse model for a test pixel by sampling similar pixels from its corresponding superpixels combinations. Firstly, multi-layer superpixels are extracted on the false color image of the HSI data by principal components analysis model. Secondly, a group of discriminative sampling pixels are exploited as reconstruction matrix of test pixel which can be jointly represented by the structured dictionary and recovered sparse coefficients. Thirdly, the orthogonal matching pursuit strategy is employed for estimating sparse vector for the test pixel. In each iteration, the approximation can be computed from the dictionary and corresponding sparse vector. Finally, the class label of test pixel can be directly determined with minimum reconstruction error between the reconstruction matrix and its approximation. The advantages of this algorithm lie in the development of complete neighborhood and homogeneous pixels to share a common sparsity pattern, and it is able to achieve more flexible joint sparse coding of spectral-spatial information. Experimental results on three real hyperspectral datasets show that the proposed joint sparse model can achieve better performance than a series of excellent sparse classification methods and superpixels-based classification methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.5
• /
• pp.2468-2483
• /
• 2017

The construction of completely random sensing matrices of Compressive Sensing requires a large number of random numbers while that of deterministic sensing operators often needs complex mathematical operations. Thus both of them have difficulty in acquiring large signals efficiently. This paper focuses on the enhancement of the practicability of the structurally random matrices and proposes a semi-deterministic sensing matrix called Partial Kronecker product of Identity and Hadamard (PKIH) matrix. The proposed matrix can be viewed as a sub matrix of a well-structured, sparse, and orthogonal matrix. Only the row index is selected at random and the positions of the entries of each row are determined by a deterministic sequence. Therefore, the PKIH significantly decreases the requirement of random numbers, which has a complex generating algorithm, in matrix construction and further reduces the complexity of sampling. Besides, in order to process large signals, the corresponding fast sampling algorithm is developed, which can be easily parallelized and realized in hardware. Simulation results illustrate that the proposed sensing matrix maintains almost the same performance but with at least 50% less random numbers comparing with the popular sampling matrices. Meanwhile, it saved roughly 15%-35% processing time in comparison to that of the SRM matrices.

• Journal of the Korean Society of Radiology
• /
• v.11 no.7
• /
• pp.655-661
• /
• 2017

Sparse view CT has been widely used to reduce radiation dose to patient in radiation therapy. In this work, we performed sinogram restoration from sparse sampling data by using inpainting method for simulation and experiment. Sinogram restoration was performed in accordance with sampling angle and restoration method, and their results were validated with root mean square error (RMSE) and image profiles. Simulation and experiment are designed to fan beam scan for various projection angles. Sparse data in sinogram were restored by using linear interpolation and inpainting method. Then, the restored sinogram was reconstructed with filtered backprojection (FBP) algorithm. The results showed that RMSE and image profiles were depended on the projection angles and restoration method. Based on the simulation and experiment, we found that inpainting method could be improved for sinogram restoration in comparison to linear interpolation method for estimating RMSE and image profiles.

• Progress in Medical Physics
• /
• v.27 no.3
• /
• pp.105-110
• /
• 2016

Sparse angular sampling has been studied recently owing to its potential to decrease the radiation exposure from computed tomography (CT). In this study, we investigated the analytic reconstruction algorithm in sparse angular sampling using the sinogram interpolation method for improving image quality and computation speed. A prototype of the spectral CT system, which has a 64-pixel Cadmium Zinc Telluride (CZT)-based photon-counting detector, was used. The source-to-detector distance and the source-to-center of rotation distance were 1,200 and 1,015 mm, respectively. Two energy bins (23~33 keV and 34~44 keV) were set to obtain two reconstruction images. We used a PMMA phantom with height and radius of 50.0 mm and 17.5 mm, respectively. The phantom contained iodine, gadolinium, calcification, and lipid. The Feld-kamp-Davis-Kress (FDK) with the sinogram interpolation method and Maximum Likelihood Expectation Maximization (MLEM) algorithm were used to reconstruct the images. We evaluated the signal-to-noise ratio (SNR) of the materials. The SNRs of iodine, calcification, and liquid lipid were increased by 167.03%, 157.93%, and 41.77%, respectively, with the 23~33 keV energy bin using the sinogram interpolation method. The SNRs of iodine, calcification, and liquid state lipid were also increased by 107.01%, 13.58%, and 27.39%, respectively, with the 34~44 keV energy bin using the sinogram interpolation method. Although the FDK algorithm with the sinogram interpolation did not produce better results than the MLEM algorithm, it did result in comparable image quality to that of the MLEM algorithm. We believe that the sinogram interpolation method can be applied in various reconstruction studies using the analytic reconstruction algorithm. Therefore, the sinogram interpolation method can improve the image quality in sparse-angular sampling and be applied to CT applications.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.5
• /
• pp.2141-2155
• /
• 2020

In order to meet the requirements of background change, illumination variation, moving shadow interference and high accuracy in object detection of moving camera, and strive for real-time and high efficiency, this paper presents an object detection algorithm based on sparse approximation recursion and sparse coding migration in subspace. First, low-rank sparse decomposition is used to reduce the dimension of the data. Combining with dictionary sparse representation, the computational model is established by the recursive formula of sparse approximation with the video sequences taken as subspace sets. And the moving object is calculated by the background difference method, which effectively reduces the computational complexity and running time. According to the idea of sparse coding migration, the above operations are carried out in the down-sampling space to further reduce the requirements of computational complexity and memory storage, and this will be adapt to multi-scale target objects and overcome the impact of large anomaly areas. Finally, experiments are carried out on VDAO datasets containing 59 sets of videos. The experimental results show that the algorithm can detect moving object effectively in the moving camera with uniform speed, not only in terms of low computational complexity but also in terms of low storage requirements, so that our proposed algorithm is suitable for detection systems with high real-time requirements.

• Journal of Digital Contents Society
• /
• v.16 no.4
• /
• pp.605-613
• /
• 2015

This paper addresses the problem to make sparse the projection matrix in pattern recognition method. Recently, the size of computer program is often restricted in embedded systems. It is very often that developed programs include some constant data. For example, many pattern recognition programs use the projection matrix for dimension reduction. To improve the recognition performance, very high dimensional feature vectors are often extracted. In this case, the projection matrix can be very big. Recently, RSR(roated sparse regression) method[1] was proposed. This method has been proved one of the best algorithm that obtains the sparse matrix. We propose three methods to improve the RSR; outlier removal, sampling and elastic net RSR(E-RSR) in which the penalty term in RSR optimization function is replaced by that of the elastic net regression. The experimental results show that the proposed methods are very effective and improve the sparsity rate dramatically without sacrificing the recognition rate compared to the original RSR method.

• Current Optics and Photonics
• /
• v.6 no.3
• /
• pp.288-296
• /
• 2022

We propose and experimentally demonstrate a novel photonic compressive sensing (CS) scheme for acquiring sparse radio frequency signals with adjustable compression ratio in this paper. The sparse signal to be measured and a pseudo-random binary sequence are modulated on consecutively connected chirped pulses. The modulated pulses are compressed into short pulses after propagating through a dispersive element. A programmable optical filter based on spatial light modulator is used to realize spectral segmentation and demultiplexing. After spectral segmentation, the compressed pulses are transformed into several sub-pulses and each of them corresponds to a measurement in CS. The major advantage of the proposed scheme lies in its adjustable compression ratio, which enables the system adaptive to the sparse signals with variable sparsity levels and bandwidths. Experimental demonstration and further simulation results are presented to verify the feasibility and potential of the approach.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.5
• /
• pp.25-31
• /
• 2009

The recently developed sampling theory, "compressed sensing" is gathering huge interest in MR reconstruction area because of its feasibility of high spatio-temporal resolution of dynamic MRI which has been limited in conventional methods based on Nyquist sampling theory. Since dynamic MRI usually has high redundant information along temporal direction, this can be very sparsely represented in most of cases. Therefore, compressed sensing that exploits the sparsity of unknown images can be effectively applied in most of dynamic MRI. This review article briefly introduces currently proposed compressed sensing based dynamic MR imaging algorithms and other methods exploiting sparsity. By comparing them with conventional methods, you may have insight how the compressed sensing based methods can impact nearly every area of clinical dynamic MRI.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.86-89
• /
• 2003

To model a numerical problem space under the limitation of available data, we need to extract sparse but key points from the space and to efficiently approximate the space with them. This study proposes a sampling method based on the search process of genetic algorithm and a space modeling method based on least-squares approximation using the summation of Gaussian functions. We conducted simulations to evaluate them for several kinds of problem spaces: DeJong's, Schaffer's, and our original one. We then compared the performance between our sampling method and sampling at regular intervals and that between our modeling method and modeling using a polynomial. The results showed that the error between a problem space and its model was the smallest for the combination of our sampling and modeling methods for many problem spaces when the number of samples was considerably small.