Aamir Saeed MALIK Tae-Sun CHOI
A classification method is presented for differentiating honeycombed High Resolution Computed Tomographic (HRCT) images from normal HRCT images. For successful classification of honeycombed HRCT images, a complete set of methods and algorithms is described from segmentation to extraction to feature selection to classification. Wavelet energy is selected as a feature for classification using K-means clustering. Test data of 20 patients are used to validate the method.
Hengliang ZHU Xuan ZENG Xu LUO Wei CAI
For variation-aware capacitance extraction, stochastic collocation method (SCM) based on Homogeneous Chaos expansion has the exponential convergence rate for Gaussian geometric variations, and is considered as the optimal solution using a quadratic model to model the parasitic capacitances. However, when geometric variations are measured from the real test chip, they are not necessarily Gaussian, which will significantly compromise the exponential convergence property of SCM. In order to pursue the exponential convergence, in this paper, a generalized stochastic collocation method (gSCM) based on generalized Polynomial Chaos (gPC) expansion and generalized Sparse Grid quadrature is proposed for variation-aware capacitance extraction that further considers the arbitrary random probability of real geometric variations. Additionally, a recycling technique based on Minimum Spanning Tree (MST) structure is proposed to reduce the computation cost at each collocation point, for not only "recycling" the initial value, but also "recycling" the preconditioning matrix. The exponential convergence of the proposed gSCM is clearly shown in the numerical results for the geometric variations with arbitrary random probability.
Congyan LANG De XU Shuoyan LIU Ning LI
Salient Region Extraction provides an alternative methodology to image description in many applications such as adaptive content delivery and image retrieval. In this paper, we propose a robust approach to extracting the salient region based on bottom-up visual attention. The main contributions are twofold: 1) Instead of the feature parallel integration, the proposed saliencies are derived by serial processing between texture and color features. Hence, the proposed approach intrinsically provides an alternative methodology to model attention with low implementation complexity. 2) A constructive approach is proposed for rendering an image by a non-linear intensity mapping, which can efficiently eliminate high contrast noise regions in the image. And then the salient map can be robustly generated for a variety of nature images. Experiments show that the proposed algorithm is effective and can characterize the human perception well.
Kenichi YABUTA Hitoshi KITAZAWA Toshihisa TANAKA
Because of an increasing number of security cameras, it is crucial to establish a system that protects the privacy of objects in the recorded images. To this end, we propose a framework of image processing and data hiding for security monitoring and privacy protection. First, we state the requirements of the proposed monitoring systems and suggest possible implementation that satisfies those requirements. The underlying concept of our proposed framework is as follows: (1) in the recorded images, the objects whose privacy should be protected are deteriorated by appropriate image processing; (2) the original objects are encrypted and watermarked into the output image, which is encoded using an image compression standard; (3) real-time processing is performed such that no future frame is required to generate on output bitstream. It should be noted that in this framework, anyone can observe the decoded image that includes the deteriorated objects that are unrecognizable or invisible. On the other hand, for crime investigation, this system allows a limited number of users to observe the original objects by using a special viewer that decrypts and decodes the watermarked objects with a decoding password. Moreover, the special viewer allows us to select the objects to be decoded and displayed. We provide an implementation example, experimental results, and performance evaluations to support our proposed framework.
Nonnegative matrix factorization (NMF) and its extensions such as Nonnegative Tensor Factorization (NTF) have become prominent techniques for blind sources separation (BSS), analysis of image databases, data mining and other information retrieval and clustering applications. In this paper we propose a family of efficient algorithms for NMF/NTF, as well as sparse nonnegative coding and representation, that has many potential applications in computational neuroscience, multi-sensory processing, compressed sensing and multidimensional data analysis. We have developed a class of optimized local algorithms which are referred to as Hierarchical Alternating Least Squares (HALS) algorithms. For these purposes, we have performed sequential constrained minimization on a set of squared Euclidean distances. We then extend this approach to robust cost functions using the alpha and beta divergences and derive flexible update rules. Our algorithms are locally stable and work well for NMF-based blind source separation (BSS) not only for the over-determined case but also for an under-determined (over-complete) case (i.e., for a system which has less sensors than sources) if data are sufficiently sparse. The NMF learning rules are extended and generalized for N-th order nonnegative tensor factorization (NTF). Moreover, these algorithms can be tuned to different noise statistics by adjusting a single parameter. Extensive experimental results confirm the accuracy and computational performance of the developed algorithms, especially, with usage of multi-layer hierarchical NMF approach [3].
Takeyuki TAMURA Tatsuya AKUTSU
The Boolean network (BN) is a mathematical model of genetic networks. It is known that detecting a singleton attractor, which is also called a fixed point, is NP-hard even for AND/OR BNs (i.e., BNs consisting of AND/OR nodes), where singleton attractors correspond to steady states. Though a naive algorithm can detect a singleton attractor for an AND/OR BN in O(n 2n) time, no O((2-ε)n) (ε > 0) time algorithm was known even for an AND/OR BN with non-restricted indegree, where n is the number of nodes in a BN. In this paper, we present an O(1.787n) time algorithm for detecting a singleton attractor of a given AND/OR BN, along with related results. We also show that detection of a singleton attractor in a BN with maximum indegree two is NP-hard and can be polynomially reduced to a satisfiability problem.
JiYing WU QiuQi RUAN Gaoyun AN
A novel bimodal method for face recognition under low-level lighting conditions is proposed. It fuses an enhanced gray level image and an illumination-invariant geometric image at the feature-level. To further improve the recognition performance under large variations in attributions such as poses and expressions, discriminant features are extracted from source images using the wavelet transform-based method. Features are adaptively fused to reconstruct the final face sample. Then FLD is used to generate a supervised discriminant space for the classification task. Experiments show that the bimodal method outperforms conventional methods under complex conditions.
Muhammad Anwaar MANZAR Tanweer Ahmad CHEEMA Abdul JALIL Ijaz Mansoor QURESHI
Image matching is an important area of research in the field of artificial intelligence, machine vision and visual navigation. This paper presents a new image matching scheme suitable for visual navigation. In this scheme, gray scale images are sliced and quantized to form sub-band binary images. The information in the binary images is then signaturized to form a vector space and the signatures are sorted as per significance. These sorted signatures are then normalized to transform the represented image pictorial features in a rotation and scale invariant form. For the image matching these two vector spaces from both the images are compared in the transformed domain. This comparison yields efficient results directly in the image spatial domain avoiding the need of image inverse transformation. As compared to the conventional correlation, this comparison avoids the wide range of square error calculations all over the image. In fact, it directly guides the solution to converge towards the estimate given by the adaptive prediction for a high speed performance in an aerial video sequence. A four dimensional solution population scheme has also been presented with a matching confidence factor. This factor helps in terminating the iterations when the essential matching conditions have been achieved. The proposed scheme gives robust and fast results for normal, scaled and rotated templates. Speed comparison with older techniques shows the computational viability of this new technique and its much lesser dependence on image size. The method also shows noise immunity at 30 dB AWGN and impulsive noise.
Seiji HAYASHI Hiroyuki INUKAI Masahiro SUGUIMOTO
The present paper describes quality enhancement of speech corrupted by an additive background noise in a single-channel system. The proposed approach is based on the introduction of a perceptual criterion using a frequency-weighting filter in a subtractive-type enhancement process. Although this subtractive-type method is very attractive because of its simplicity, it produces an unnatural and unpleasant residual noise. Thus, it is difficult to select fixed optimized parameters for all speech and noise conditions. A new and effective algorithm is thus developed based on the masking properties of the human ear. This newly developed algorithm allows for an automatic adaptation in the time and frequency of the enhancement system and determines a suitable noise estimate according to the frequency of the noisy input speech. Experimental results demonstrate that the proposed approach can efficiently remove additive noise related to various kinds of noise corruption.
Kazumi SAITO Takeshi YAMADA Kazuhiro KAZAMA
To understand the structural and functional properties of large-scale complex networks, it is crucial to efficiently extract a set of cohesive subnetworks as communities. There have been proposed several such community extraction methods in the literature, including the classical k-core decomposition method and, more recently, the k-clique based community extraction method. The k-core method, although computationally efficient, is often not powerful enough for uncovering a detailed community structure and it produces only coarse-grained and loosely connected communities. The k-clique method, on the other hand, can extract fine-grained and tightly connected communities but requires a substantial amount of computational load for large-scale complex networks. In this paper, we present a new notion of a subnetwork called k-dense, and propose an efficient algorithm for extracting k-dense communities. We applied our method to the three different types of networks assembled from real data, namely, from blog trackbacks, word associations and Wikipedia references, and demonstrated that the k-dense method could extract communities almost as efficiently as the k-core method, while the qualities of the extracted communities are comparable to those obtained by the k-clique method.
Chin-Feng TSAI Huan-Sheng WANG King-Chu HUNG Shih-Chang HSIA
Wavelet-based features with simplicity and high efficacy have been used in many pattern recognition (PR) applications. These features are usually generated from the wavelet coefficients of coarse levels (i.e., high octaves) in the discrete periodized wavelet transform (DPWT). In this paper, a new 1-D non-recursive DPWT (NRDPWT) is presented for real-time high octave decomposition. The new 1-D NRDPWT referred to as the 1-D RRO-NRDPWT can overcome the word-length-growth (WLG) effect based on two strategies, resisting error propagation and applying a reversible round-off linear transformation (RROLT) theorem. Finite precision performance analysis is also taken to study the word length suppression efficiency and the feature efficacy in breast lesion classification on ultrasonic images. For the realization of high octave decomposition, a segment accumulation algorithm (SAA) is also presented. The SAA is a new folding technique that can reduce multipliers and adders dramatically without the cost of increasing latency.
Chien-Tsun CHEN Yu Chin CHENG Chin-Yun HSIEH
Design by Contract (DBC), originated in the Eiffel programming language, is generally accepted as a practical method for building reliable software. Currently, however, few languages have built-in support for it. In recent years, several methods have been proposed to support DBC in Java. We compare eleven DBC tools for Java by analyzing their impact on the developer's programming activities, which are characterized by seven quality attributes identified in this paper. It is shown that each of the existing tools fails to achieve some of the quality attributes. This motivates us to develop ezContract, an open source DBC tool for Java that achieves all of the seven quality attributes. ezContract achieves streamlined integration with the working environment. Notably, standard Java language is used and advanced IDE features that work for standard Java programs can also work for the contract-enabled programs. Such features include incremental compilation, automatic refactoring, and code assist.
Ryoji HASHIMOTO Tomoya MATSUMURA Yoshihiro NOZATO Kenji WATANABE Takao ONOYE
A multi-agent object attention system is proposed, which is based on biologically inspired attractor selection model. Object attention is facilitated by using a video sequence and a depth map obtained through a compound-eye image sensor TOMBO. Robustness of the multi-agent system over environmental changes is enhanced by utilizing the biological model of adaptive response by attractor selection. To implement the proposed system, an efficient VLSI architecture is employed with reducing enormous computational costs and memory accesses required for depth map processing and multi-agent attractor selection process. According to the FPGA implementation result of the proposed object attention system, which is accomplished by using 7,063 slices, 640512 pixel input images can be processed in real-time with three agents at a rate of 9 fps in 48 MHz operation.
This paper presents a new adaptive minor subspace extraction algorithm based on an idea of Peng and Yi ('07) for approximating the single minor eigenvector of a covariance matrix. By utilizing the idea inductively in the nested orthogonal complement subspaces, the proposed algorithm succeeds to relax the numerical sensitivity which has been annoying conventional adaptive minor subspace extraction algorithms for example, Oja algorithm ('82) and its stabilized version: O-Oja algorithm ('02). Simulation results demonstrate that the proposed algorithm realizes more stable convergence than O-Oja algorithm.
Thinning and line extraction of binary images not only reduces data storage amount, automatically creates the adjacency and relativity between line and points but also provides applications for automatic inspection systems, pattern recognition systems and vectorization. Based on the features of construction drawings, new thinning and line extraction algorithms were proposed in this study. The experimental results showed that the proposed method has a higher reliability and produces better quality than the various existing methods.
This paper presents a closed form solution to a problem of constructing the best lower bound of a convex function under certain conditions. The function is assumed (I) bounded below by -ρ, and (II) differentiable and its derivative is Lipschitz continuous with Lipschitz constant L. To construct the lower bound, it is also assumed that we can use the values ρ and L together with the values of the function and its derivative at one specified point. By using the proposed lower bound, we derive a computationally efficient deep monotone approximation operator to the level set of the function. This operator realizes better approximation than subgradient projection which has been utilized, as a monotone approximation operator to level sets of differentiable convex functions as well as nonsmooth convex functions. Therefore, by using the proposed operator, we can improve many signal processing algorithms essentially based on the subgradient projection.
Daisuke ABE Eigo SEGAWA Osafumi NAKAYAMA Morito SHIOHARA Shigeru SASAKI Nobuyuki SUGANO Hajime KANNO
In this paper, we present a robust small-object detection method, which we call "Frequency Pattern Emphasis Subtraction (FPES)", for wide-area surveillance such as that of harbors, rivers, and plant premises. For achieving robust detection under changes in environmental conditions, such as illuminance level, weather, and camera vibration, our method distinguishes target objects from background and noise based on the differences in frequency components between them. The evaluation results demonstrate that our method detected more than 95% of target objects in the images of large surveillance areas ranging from 30-75 meters at their center.
Haruna MATSUSHITA Yoshifumi NISHIO
In the real world, it is not always true that neighboring houses are physically adjacent or close to each other. in other words, "neighbors" are not always "true neighbors." In this study, we propose a new Self-Organizing Map (SOM) algorithm, SOM with False-Neighbor degree between neurons (called FN-SOM). The behavior of FN-SOM is investigated with learning for various input data. We confirm that FN-SOM can obtain a more effective map reflecting the distribution state of input data than the conventional SOM and Growing Grid.
Xiao WU Ming LI Hongbin SUO Yonghong YAN
In this letter we focus on the task of selecting the melody track from a polyphonic MIDI file. Based on the intuition that music and language are similar in many aspects, we solve the selection problem by introducing an n-gram language model to learn the melody co-occurrence patterns in a statistical manner and determine the melodic degree of a given MIDI track. Furthermore, we propose the idea of using background model and posterior probability criteria to make modeling more discriminative. In the evaluation, the achieved 81.6% correct rate indicates the feasibility of our approach.
Hiroyuki SAKAI Shigeru MASUYAMA
We propose a method of extracting cause information from Japanese financial articles concerning business performance. Our method acquires cause information, e.g. "(zidousya no uriage ga koutyou: Sales of cars were good)". Cause information is useful for investors in selecting companies to invest. Our method extracts cause information as a form of causal expression by using statistical information and initial clue expressions automatically. Our method can extract causal expressions without predetermined patterns or complex rules given by hand, and is expected to be applied to other tasks for acquiring phrases that have a particular meaning not limited to cause information. We compared our method with our previous one originally proposed for extracting phrases concerning traffic accident causes and experimental results showed that our new method outperforms our previous one.