Earticle

In the field of computer vision and photogrammetry, it is constantly necessary to online or real-time acquire the camera parameters through self-calibration. This paper presents a method about self-calibration of camera with rotary motion based on SIFT feature matching. The proposed approach first shoots more than three images of the same scene by rotating the camera and keeping its position and internal parameters unchanging. After SIFT feature extraction and sequentially cycled matching for all images, the optimal reference image and effective images are determined by virtue of the algorithm on pose estimation. According to the coordinates of matched SIFT features, all 2D projection transformation matrices which transforms the reference into other effective images are calculated. With these matrices, relevant linear equations are established and the internal matrix of camera is solved. The proposed method can be online applied to quickly, accurately and stably obtain internal parameters of camera. Real data has been used to test the proposed approach, and very good results have been achieved.

In this paper, we propose an efficient algorithm for encoding of MPEG-4 simple profile. In the proposed algorithm, I frame encoding adopts all phase biorthogonal transform (APBT); for P frames, each frame puts the reconstructed frame of previous coded one as the reference frame to perform motion estimation (ME) and motion compensation (MC), and then the encoding of residual frame uses discrete cosine transform (DCT). While the proposed algorithm applying on MPEG-4, the objective effect and visual quality of reconstructed video sequence are approximate to the one based on DCT at the same bit rates. But the advantage is that the APBT transform coefficients can be quantized uniformly. Therefore the computing time is shorted and hardware implementation becomes easier. The proposed algorithm uses low computational complexity to make it applicable to encoding of MPEG-4 simple profile.

The use of electronic documents has rapidly increased in recent decades and the PDF is one the most commonly used electronic document formats. A scanned PDF is an image and does not actually contain any text. For the vision–impaired user who is dependent upon a screen reader to access this information, this format is not useful. Thus addressing PDF accessibility through assistive technology has now become an important concern. PDF layout analysis provides precious formatting information that supports PDF component classification. This classification facilitates the tag generation. Accurate tagging produces a searchable and navigable scanned PDF document. This paper describes several practical segmentation methods which are easy to implement and efficient for PDF layout analysis so that the scanned PDF document can be navigated or searched using assistive technologies.

Stereo matching is one of the key technologies in stereo vision system due to its ultra high data bandwidth requirement, heavy memory accessing and algorithm complexity. To speed up stereo matching, various algorithms are implemented by different software and hardware processing methods. This paper presents a survey of stereo matching software and hardware implementation research status based on local and global algorithm analysis. Based on different processing platforms, including CPU, DSP, GPU, FPGA and ASIC, analysis are made on software or hardware realization performance, which is represented by frame rate, efficiency represented by MDES, and processing quality represented by error rate. Among them, GPU, FPGA and ASIC implementations are suitable for real-time embedded stereo matching applications, because they are low power consumption, low cost, and have high performance. Finally, further stereo matching optimization technologies are pointed out, including both algorithm and parallelism optimization for data bandwidth reduction and memory storage strategy.

Nowadays, for providing the secure facilities and services to the user, the accurate identification is necessary. The Iris recognition is one of the attractive approach for user’s identification, provides high level of security and convenience then the other methods of identification like traditional ID and password, which can be lost or transferred. However, the iris recognition algorithms are implemented on general purpose sequential processing systems, such as generic central processing units (CPUs). Parallel processing is an alternative offers an opportunity to enhance the performance of system by increasing the speed. The most time consuming part of Iris recognition algorithm is matching part, which is implemented using Verilog HDL through ISE Design suit (14.2), achieved significantly reduction in execution time. The proposed design is suitable for integration either in ASIC or FPGA.

To achieve the color consistent reproduction of multispectral images in the different viewing condition, a new method of multispectral image reproduction via color appearance mapping was proposed. Firstly, through the introduction of color appearance transformation, the color appearance description of the source spectral reflectance in the source viewing condition was obtained. Then by the construction of the inverse model, the high dimension spectra in the destination viewing condition were evaluated, which was color appearance matching but spectral mismatching with the source image. Finally, to improve the spectral precision of the reproduced spectra, the evaluated spectra were corrected by the method of metamerism correction based on the source spectra, and then the reproduced spectral image was obtained, which matched the source image in color appearance and in spectra when the reproduction viewing condition was different from the source. Experiments show that the perceptual color difference and the spectral error between the reproduced multi-spectral image and its original in the different viewing condition are small. The new method preserves the spectral information of the source multispectral image and achieves equal perceptual reproduction to the source image.

Texture harmony pursues a suitable texture matching to meet the customers Kansei image requirements. The texture harmony method which is based on Kansei Engineer is developed. Several questionnaires are made to obtain the Kansei words, design elements and texture factors. The representative Kansei words, representative design elements and the representative texture factors are selected using Pareto Diagram, Likert scale, multidimensional scaling analysis and cluster analysis. After developing the virtual samples, the respondents are asked to evaluate the Kansei image score of each sample according to the different Kansei image word. The Kansei image evaluating matrix is obtained by combining the Kansei image score with the texture combination code. The multiple linear regression model is supposed to explain the relationship of the Kansei image score and the texture factors. Based on the Kansei image evaluate matrix, the hypothesis is verified using the SPSS software. The case of electric kettle texture harmony design is studied to verify the method. The method can facilitate designers work, and lay a foundation of computer aided texture design system.

This paper aims to obtain an integrated and consecutive original image from noisy image by a directional correlation-dependent filtering technique, for removing salt and pepper noise from corrupted images. A comparative study is made on the various image denoising methods and extensive experimental results demonstrate that directional correlation-dependent filtering algorithm can obtain better performances in terms of PSNR value compare to other impulse denoising techniques. The proposed algorithm can preserve edges very well while removing impulse noise and is very suitable to be applied to many real-time applications.

Traditional radar target recognition algorithms utilize balance data set to train the classifier and achieve a satisfactory result on a balance test data set. However, in the case of non-cooperative target recognition, we only obtain a small amount of non-cooperative target samples, while we can obtain a larger number of cooperative target samples easily, which leads to an imbalance training data set. In this paper, we consider the imbalance data classification problem in radar target recognition. We utilize the cost-sensitive approach and assume that different kinds of mistakes lead to different losses. Based on this assumption, a novel radar target recognition algorithm, called isometric cost-sensitive Laplacian eigenmaps (ICSLE), is presented. The basic idea of ICSLE is that the larger the misclassification cost is, the further the distance between two classes is, and vice versa. Moreover, in order to effectively utilize the cost information and local property of observation samples, we use the geodesic distance as the edge weight, instead of the local Euclidean distance. Experiments on millimeter wave radar high-resolution range profile (HRRP) demonstrate the effectiveness of our method.

Nowadays the content based image retrieval (CBIR) is becoming a source of exact and fast retrieval. CBIR presents challenges in indexing, accessing of image data and how end systems are evaluated. Data clustering is an unsupervised method for extraction hidden pattern from huge data sets. Many clustering and segmentation algorithms both suffer from the limitation of the number of clusters specified by a human user. It is often impractical to expect a human with sufficient domain knowledge to be available to select the number of clusters (NC) to return. This paper discusses the image retrieval based on NC which is evaluated using hierarchical agglomerative clustering algorithm (HAC). In this paper, we determine the optimal number of clusters using HAC applied on RGB images and validate them using some validity indices. Based on number of clusters, we retrieve set of images. These cluster values can be further used for divide and conquer technology and indexing for large image dataset. An experimental study is presented on real data sets.

Abnormal event detection in nature settings is an active issue in computer vision domain. A novel unsupervised method is proposed to detect abnormal events by combining dynamic texture and sparse coding. In this method, dynamic texture is used as descriptors in a spatio-temporal manner to describe spatio-temporal volumes of events in videos. Sparse coding is utilized for reconstructing the testing data to measure its normalness. Experiments are conducted on the well known UCSD dataset and UMN dataset to demonstrate the efficiency of the proposed method. The results show that the proposed method outperforms the current state-of-the-art methods.

Decision-making process refers in a particular situation, and choosing from several existing optional strategy evaluation process of certain strategy, however, in a complex environment, individual decisions tend to be affected by other individuals and random change of strategy, thus it is difficult to make a best decision. Although game theory provides a variety of decision-making plan, but it is not clear how the decision is influenced by the experience. The best decision-making behaviors need flexible adjustment strategy accord to the recent gains. In order to explore the nature of the human decision-making process USES the zero-sum game COINS of the positive and negative game, assuming that flexibility was produced in the process of reinforcement learning and a reinforcement learning model is established. In addition, the entire record the event related potential in the process of the subjects in the experiment. Analysis of ERP (Event Related Potential) data mainly focus on feedback related negative wave, which was considered to reflect the error signal and benefit related to the brain potential. Results show that after the loss to the other party of the event related potential amplitude suggested that participants would change the policy of pushing in the subsequent regulations, and the result was satisfied.

The image processing is very important in several applications and have been using in them very efficiently. Normally all the pixels in every image are in shape of square grid but most of the time, the feature extraction from an image like image segmentation, image detection, edge detection, texture recognition etc becomes difficult to recognize in square pixel images. So one new approach called hexagonal pixel structure; has been designed to overcome the problems of square pixels structure. This paper gives an overview of different square pixels to hexagonal pixels representation techniques.

The noise processing is the key of improving recognition rate for the noisy utterance. While for the short utterance, its corpus is less and small amount of speech data is available for testing and training, so making full use of its limit corpus is the key of improving recognition rate of the short utterance. For the noisy short utterance, the noise processing and making full use of the limit corpus are vital. We proposed noise separation algorithm based on constrained Non-negative matrix factorization (CNMF) to make the noise processing. As making full use of the limit corpus, we proposed the improved SNR discrimination algorithm (ISNRDA) and the differences detection and discrimination algorithm (DDADA), we use the two classification algorithm to estimate the quality of the speech frame, and classify the speech frame. Besides, we combine the above classification result with the GMM-UBM three-stage classification model proposed in this paper, so that we can make full use of the limit corpus of the noisy short utterance. Experiments show that the above algorithms can improve speaker recognition performance of noisy short utterance.

A generalized S transform based on energy normalized windows is presented, which avoids the weighting effect of existing S transform influenced by the time-frequency spectral energy. The transform can reflect the spectral energy distribution of different frequency components while maintaining the excellent time-frequency resolution. While TT transform is a new approach to process non-stationary signal based on S transform, which has differential concentration for different frequency components, and lower frequency components can be suppressed through extracting the TT spectrum’s diagonal elements. However, this method has a deficiency that part of high frequency interference is retained during processing low frequency signal. Therefore, time-frequency filtering based on the generalizing S transform is used to suppress the high frequency interference and TT transform is adopted to suppress the low frequency contents. Finally, this method is used to process seismic data, the result demonstrates the effective of combining energy normalized S transform and TT transform.

Image Registration is the process of aligning two or more images of the same scene taken from different viewpoints, at different times, from different sensors. Image registration aligns two images, i.e., base image and reference image, geometrically. There are different approaches of image registration and these approaches are categorized according to their nature that is areas based and feature based. These approaches are also categorized according to the four simple steps of image registration procedure: feature detection, feature matching, function mapping and transformation and re-sampling. Advantages and disadvantages of different methods are discussed in the paper. The main aim of this paper is to provide the knowledge of different image registrations methods used in different application area.

Concrete CT image segmentation is a hot civil engineering research field in recent years. An algorithm of minimizing the energy of local region is proposed to solve the problem of interaction fails to capture statistical property of natural image and to segment concrete CT image. At first, the algorithm utilizes local region information of image to construct the energy model of local region, establishes a segmentation model of local interaction region based on MRF. Then using loopy belief propagation (LBP) algorithm and other algorithms optimizes global energy. In process of optimization, local region energy is converged, and the label of local region is estimated based on MAP algorithm. Then the information of local region is transferred to the region of neighborhood. The result of experiment shows that comparing with standard LBP algorithm, the new algorithm has a better segmentation result, and efficiently restrained effect on image noise and texture for segmentation. Result of concrete CT image segmentation will simplify the following CT statistical analysis and provide an important method for reaching real meso-structure of concrete’s finite element network.

Brain image segmentation is an important part of medical image analysis. Due to the effect of imaging mechanism, MR images usually intensity in homogeneity, which is also named as bias field. Traditional Gaussian Mixed Model (GMM) method is hard to obtain satisfied segmentation results with the effect of noise and bias field. We propose a novel model based on GMM and nonlocal information. The improved method coupled segmentation and bias field correction that can manage the bias field while segmenting the image. In order to obtain a smooth bias field, we employed the Legendre Polynomials to fit it and merged it to the EM framework. We also use the non local information to deal with the noise and preserve geometrical edges information. The results show that our method can obtain more accurate results and bias field.

This article provides a comparison between methods of images amplification with good results in the preservation of the structural similarity of the output image. This parameter is evaluated through the method SSIM (structural similarity index).One of the methods used for amplification is the combination of SWT (stationary wavelet transform) for the decomposition of the image, SVD (singular value decomposition) for conservation of high frequency content and DWT ( discrete Wavelet Transform ) for obtain the high resolution image. This method has report of good results [5]. Other amplification technique is proposed, RHPI (pulses interpolation constrained by high frequency)[9].The procedure RHPI uses a theoretical model of digital image registration and high frequency filters for the construction of the interpolation kernel for the amplification process. Finally, both methods are compared through SSIM structural similarity coefficient and PSNR parameters. The proposed method RHPI is a solution better than the SWT-SVD-DWT for the improvement in the conservation of SSIM and PSNR ratios for high rates of amplification.

Image segmentation successfully achieved by employing image binarization technique. In this paper, a new method is proposed for detecting the vehicle logo in an image. The proposed method is applied to determine a local threshold for each pixel. In this method, we use the different value of lambda. In our proposed method, we use the double mean filter to remove noise. Experimental resultsproofthatthe proposed methodisfast and successfully detect all edges,with minornoise.In addition, it has the feature of smaller entropycompared toothermethods.

In the interest of deriving regressor that is robust to outliers, we propose a support vector regression (SVR) based on non-convex quadratic insensitive loss function with flexible coefficient and margin. The proposed loss function can be approximated by a difference of convex functions (DC). The resultant optimization is a DC program. We employ Newton’s method to solve it. The proposed model can explicitly enhance the robustness and sparseness of SVR. Numerical experiments on six benchmark data sets show that it yields promising results.

In this paper, we have proposed an optimized sparse representation algorithm based on Log-Gabor (Sparse Representation-based Classification Based on Log-Gabor, Log-GSRC), which applies local features information of samples to the sparse representation method. Actually, SRC (Sparse Representation-based Classification) is using a linear correlation between the samples of one class which can be assumed that these samples exist in a subspace, and also can be linear represented with each other. It is a global representation and it ignores the local features information of the samples, while in the case of there are a smaller number of training samples per class, SRC will obtain an inaccurate classification result which may correspond to one and more classes in the process of sparse decomposition. However, the Log-GSRC combines global and local features information of the samples and also improves the robustness of SRC. The experimental results clearly showed that Log-GSRC has much better performance than SRC and also has much higher recognition rates than SRC in face recognition.

Cellular Automata is a computational model used to describe the complex system through simple rules. It has been significantly applied to image processing operations. It appear as a natural tool for image processing because of the simplicity of mapping a digital image to a cellular automata and the ability of applying different image processing operations in real time. Noise filtering is considered to be an important operation of image processing. In this paper cellular automata based noise filter has been proposed for different levels of noise. The filter is also compared with some standard filters in terms of peak signal to noise ratio and structured symmetry index measure and it is found that the proposed model shows consistently better performance in terms of both the parameter.

In the design procedure of the reflectarray antennas, the fast analysis of the aperture efficiency and radiation patterns is necessary before manufacture to forecast the antennas performance proximately. The fast calculation of the aperture efficiency and radiation patterns of reflectarray antennas is presented in this paper. A more general approach is introduced to calculate the spillover efficiency of a reflectarray. Meanwhile, the taper efficiency of the reflect-array is calculated with a unified set of equations. Two different fast approaches of the radiation patterns analysis-array theory and aperture field are described, and the discrete Fourier transform is applied to speed up the computational speed. On the basis of these derivations, numerical results of a Ku band reflectarray with offset feed are presented to show the performance of these approaches, and to provide the guidelines for designing and analyzing reflect-array antennas.

Image restoration is the process of restoring or deblurring an image which has been undergone certain degradations. In this paper, we proposed a method for image restoration based on segmentation and fuzzy clustering. This Method consider the similar image pair which having common feature corresponding to degraded one in another. This proposed method firstly partition the image into specified segments and then it will use fuzzy clustering to cluster the segment based on their PSNR value and provide the segmentsthat needs to restore which will further used for accuracy of method.The performance of the system is evaluated on the basis of PSNR value. The proposed method shows the higher efficiency compared to existing methods.

This paper sets forth an adaptive timer handoff algorithm with respect to the shortcoming of fixed timer handoff algorithm. In the algorithm, the real-time handoff timer value T is calculated based on the distance di between current mobile station and original serving cell as well as three constant variables: rank, ram and p, thus changing the handoff waiting time in real time. The simulation results showed that: the handoff occurs only when the mobile station needs to move to a far position from the original serving cell in the fixed timer handoff algorithm and larger set timer value T, and the strength of signals received by the mobile station in the original serving cell is poor; however, in the adaptive timer handoff algorithm, the ping-pong effect can be eliminated, the average handoff times are reduced, the mean strength of signals at handoff is better, the handoff position concentrates rationally and the overall performance reaches to the optimal effect.

For pose-varied color face image, this paper proposed a method of facial feature location and pose estimation based on an unsupervised sphere skin model and fused facial features. Firstly, an adaptive preprocessing, an established unsupervised sphere model and the holes filling technology were presented to extract the face. Then the difference feature and different local binary pattern textural feature of human face were fused to used in the hybrid integral projection technology for localizing the facial features. Finally, poses were estimated using the geometrical distribution of facial features. Experimental results show that, the method can locate facial features and classify different pose more effectively and adaptively.

Trademark carries the prestigious values for a particular company so it is very important to distinguish it from the similar context trademark. In this research paper, we propose an efficient Trademark Image Retrieval (TIR) model which is a branch of Content Based Image Retrieval (CBIR). In the proposed system we extract the edge point of a particular image and after this the edge point are evaluated to find the corner pixel from it. For the performance evaluation of the system we use the most commonly used method namely precision-recall. From the experimental result we conclude that the TIR based on shape feature perform better and gives satisfactory result.

Image classification is one of the most fundamental and useful activities in computer vision domain. For better accuracy and executing efficiency under the circumstance of high dimensional feature descriptors in image classification, we proposes a novel framework for multi-class image classification based on kernel principal component analysis(KPCA) for feature descriptors post-processing and support vector machine (SVM) with randomized hyper-parameter optimization for classification. We produce the image feature representation by extracting pyramid histogram of visual word (PHOW) descriptors of image, then map the descriptors though additive kernels. At the third step we use KPCA for feature dimensionality reduction. Finally we classify image by SVM with randomized hyper-parameter optimization. Extensive experiments are tested on two data sets: Msrcv2, 15-Scenes. These experiments justify that (1) feature descriptors with KPCA is superior to that with PCA for dimensionality reduction;(2)SVM with randomized hyper-parameter optimization greatly saves time while keeping high accuracy.

This paper proposes a motion estimation algorithm that uses different search pattern. Motion Estimation Algorithm plays a vital role in video compression. In block-based motion estimation, search pattern with different shape or size has very important impact on search speed and distortion performance. In this paper, Octagon and square search pattern is proposed for block-based motion estimation. The proposed method contains 13 points that structures as Octagon for finding large motion vectors. In addition, Square Search Pattern with 8 points is used for finding small motion vectors. The speedup gain of the proposed method over the Unsymmetrical-cross Multi-hexagon-grid Search (UMHexS) algorithm is more striking. Experimental results substantially proved that the performance of the proposed method is better than the existing state-of-art motion estimation methods.