Showing papers in "Systems and Computers in Japan in 2007"
TL;DR: Although this interpolation method does not restore the original image, the authors confirmed through experimental results that it can provide interpolation without a feeling of oddness for images which have a high level of self-correlation.
Abstract: In this paper the authors propose a method to use interpolation to eliminate characters in only one image with telops or other text through an image interpolation method that uses the eigenspace method. Background scenes and other images have a fractal character, and often the self-correlation in the image can be assumed to be high. The authors focus on this point and represent rules for describing the image based on an eigenspace consisting of only one image that has defects. The eigenspace generated in this manner reflects the features of the image, and by using this eigenspace, image interpolation can be achieved. Although this interpolation method does not restore the original image, the authors confirmed through experimental results that it can provide interpolation without a feeling of oddness for images which have a high level of self-correlation. © 2006 Wiley Periodicals, Inc. Syst Comp Jpn, 38(1): 87– 96, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10319.
49 citations
TL;DR: It is shown that appearance of optically reconstructed surface object of the full-parallax holograms generated by the proposed method varies in a natural manner as the observation point is changed.
Abstract: In this paper we report a method of removing hidden surfaces in full-parallax computer-generated holograms. First, we examine the handling of light shielding by tilted patches in terms of wave optics, and then propose a silhouette approximation. In addition, we derive a recurrence formula to synthesize object light waves with hidden surface removal by the proposed method. We show that appearance of optically reconstructed surface object of the full-parallax holograms generated by the proposed method varies in a natural manner as the observation point is changed. This study achieves hidden surface removal in full-parallax holograms, which was believed to be possible only if vertical parallax is ignored. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(6): 53–61, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20345 Copyright © 2004 Wiley Periodicals, Inc.
21 citations
TL;DR: A simpler economical system for the 3-D pattern formation by using an ultraviolet laser and light-sensitive resins, which means that such complicated patterns as holes, screws and hollow objects can be formed by one process.
Abstract: There is no established technique for the 3-dimensional (3-D) pattern formation, and many structures cannot be formed using the conventional CAD/CAM technology We have constructed a simpler economical system for the 3-D pattern formation by using an ultraviolet laser and light-sensitive resins In this technique, the CAD data designed by a computer are sliced horizontally Using the sliced data, the laser is moved in a 2-D direction and the light-sensitive resin is irradiated Then solidified thin plates are stacked to form a pattern Using this technique, such complicated patterns as holes, screws and hollow objects, can be formed by one process Since it is not necessary to maintain the cutting tools and the cutting scraps, a completely automated process can be achieved
21 citations
TL;DR: This paper describes a method to calculate the model parameters from the edge of a traffic lane in an image of a road, after presenting a 3-D model of aRoad based on its design conditions.
Abstract: This paper proposes a method to record three-dimensional (3-D) geometry of a road from a single-eye image. Although generally it is impossible to obtain a 3-D information from a single image of a road edge, this becomes possible by adding information of the road from its design conditions. This paper describes a method to calculate the model parameters from the edge of a traffic lane in an image of a road, after presenting a 3-D model of a road based on its design conditions. In the proposed model it is assumed that the shapes of the center line of a road seen in horizontal and vertical planes consist of alternative straight lines and curves. The 3-D shape of a road using this assumption can be classified into four categories. First, the parameters for a pair of parallel straight lines connected to an adjacent curve are calculated. Then the parameters of the curve connected to the straight lines are calculated. A model having such synthesized road edges was computer-simulated, giving the maximum relative error of the parameters of 12 percent. Images of road edges photographed from a vehicle on an actual road were processed manually for comparison, giving a maximum relative error of 21 percent with reference to the design data of the road.
19 citations
TL;DR: Simulation-based evaluation reveals that one of the strategies, which carefully accepts newly discovered Web servers, shows superior results in terms of harvest ratescoverage and runtime efficiency.
Abstract: Many countries have created Web archiving projects aiming at long-term preservation of Web information, which is now considered precious in cultural and social aspects. However, because of its borderless character, the Web poses obstacles to comprehensively gathering information originating in a specific nation or culture.This paper proposes an efficient method for selectively collecting Web pages written in a specific language. First, a linguistic graph analysis of real Web data obtained from a large crawl is conducted in order to derive a crawling guideline, which makes use of language attributes per Web server. The guideline then is formed into a few variations of link selection strategies. Simulation-based evaluation reveals that one of the strategies, which carefully accepts newly discovered Web servers, shows superior results in terms of harvest ratescoverage and runtime efficiency. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(2): 10–20, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20693
19 citations
TL;DR: In this method, multiple linear predictors are created for each image and adaptive prediction that responds to the local structure of images such as edges and textures is achieved by switching between these predictors at the block level by using quadtree segmentation.
Abstract: A highly efficient lossless encoding method for static images is proposed. In this method, multiple linear predictors are created for each image and adaptive prediction that responds to the local structure of images such as edges and textures is achieved by switching between these predictors at the block level. Furthermore, the probability density functions of the prediction errors are categorized by context modeling and modeled by generalized Gaussian functions, and adaptive arithmetic encoding of the prediction errors is performed by using probability tables that are generated for each pixel from this model. Parameters that are needed in the coding such as the prediction coefficients, the predictor selection data for each block, and the shapes of the generalized Gaussian functions are optimized by repeatedly minimizing a cost function that includes the code length of the parameters themselves in addition to the code length of the prediction errors that are calculated from the probability model above, and the parameters are then encoded separately as side data for each image. A procedure is introduced to improve prediction accuracies by using quadtree segmentation to segment the image into variable-sized blocks between which the predictor can change. Coding experiments are conducted and the proposed method is found to produce coding rates of 6 to 44p lower than the international standard JPEG-LS method, with the proposed method achieving superior coding performance that surpasses existing coding methods for all of the images used in the experiments. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(4): 1–11, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20630
16 citations
TL;DR: This paper provides new feedback pathways around the competitive layer to refer to context information of the past history of the standard SOM, and finds that this approach is robust to both the temporal elasticity and spatial displacement.
Abstract: A self-organizing map (SOM) performs a mapping of an object preserving its topological relations between input and output spaces, and also can be seen as a coordinate transformer that preserves adjacency relations. Since the standard SOM cannot deal with temporal data intrinsically, in this paper we provide new feedback pathways around the competitive layer to refer to context information of the past history. An extra output layer is added next to the competitive layer to represent secondary candidates and a quantitative measure of confidence. From the viewpoint of its structural similarities, we refer to this architecture as an Elman-type feedback SOM. In order to clarify the effectiveness of the proposed model, we then adopt a temporal signal processing task of Braille recognition. Braille is a character set for visually impaired people made up of 6 dots on a 3 by 2 grid. When Braille is read by running the fingertip over the characters, the vertical axis can be seen as providing spatial information and the horizontal axis as providing temporal information. As a result of computer simulations with partially common four city names, we confirm that the proposed model can recognize them appropriately. In addition, we find that this approach is robust to both the temporal elasticity and spatial displacement. Moreover, when we conduct an analysis with the neuro-bar model, we find that the Braille recognition task is executed based on a series of state transition along a type of pathway that has been developed during training. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(3): 62– 71, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20260
16 citations
TL;DR: A recognition method of graphs focusing on line graph images which do not include markers comprised by solid lines, dotted lines, broken lines, and dash-dot lines is proposed and shown the effectiveness of this through experimentation.
Abstract: Up to the present there has been a great deal of research related to recognition and comprehension of document images. A large amount of this research focused on text although a significant amount of information in graph images within document images is also included. If it is possible to recognize and comprehend a graph image, electronic documents can be used more efficiently. Research on recognition methods focusing on graph images within document images dealt with solid line graph images (bar graphs and markers). This paper proposes a recognition method of graphs focusing on line graph images which do not include markers comprised by solid lines, dotted lines, broken lines, and dash-dot lines. This paper also shows the effectiveness of this through experimentation. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(14): 103114, 2007; Published online in Wiley InterScience (). DOI 10.1002/scj.10615
14 citations
TL;DR: In this paper, a multiview video coding method to achieve high coding efficiency with low-delay random access functionality is proposed, and the coding method of disparity vectors, which utilizes the camera arrangement, is proposed.
Abstract: We have proposed free-viewpoint video communications, in which a viewer can change the viewpoint and viewing angle when receiving and watching video content. A free-viewpoint video consists of several views, whose viewpoints are different. To freely and instantaneously change the viewpoint and view angle, a random access capability to decode the requested view with little delay is necessary. In this paper, a multiview video coding method to achieve high coding efficiency with low-delay random access functionality is proposed. In the proposed method, the GOP is the basic unit of a view, and selective reference picture memory management is applied to multiple GOPs to improve coding efficiency. In addition, the coding method of disparity vectors, which utilizes the camera arrangement, is proposed. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(5): 14– 29, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20683
14 citations
TL;DR: Experimental evaluation results show that the proposed method of domain classification of a technical term achieved mostly 90p precisionsrecall, and the task of discovering novel technical terms that are not included in any existing lexicons of technical terms of the domain was discovered.
Abstract: This paper proposes a method of domain classification of technical terms using the Web. In the proposed method, it is assumed that, for a certain technical domain, a list of known technical terms of the domain is given. Technical documents of the domain are collected through the Web search engine, which are then used for generating a vector space model for the domain. The domain specificity of a target term is estimated according to the distribution of the domain of the sample pages of the target term. Experimental evaluation results show that the proposed method of domain classification of a technical term achieved mostly 90p precisionsrecall. We then apply this technique of estimating domain specificity of a term to the task of discovering novel technical terms that are not included in any existing lexicons of technical terms of the domain. Out of 1000 randomly selected candidates of technical terms per domain, we discovered about 100 to 200 novel technical terms. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(14): 1119, 2007; Published online in Wiley InterScience (). DOI 10.1002/scj.20852
13 citations
TL;DR: A new approach to estimating muscular contraction parameters for the purpose of generating facial expressions that can generate various facial expressions of the individual facial model, which corresponds to the facial expression of the actual face image.
Abstract: We aim to generate facial expressions of various persons more naturally and without any markers and feature lines. We propose a new approach to estimating muscular contraction parameters for the purpose of generating facial expressions. The muscles of the face are commonly known as the muscles of facial expression. The facial expressions of a human are generated by muscle movements inside the face. If the muscle-based facial model that Waters proposed is used, facial expressions can be generated by muscular contraction parameters. In this paper, first, the facial surface feature points of the face images are detected by image-processing methods. Next, the muscular contraction parameters are estimated by the neutral expression and arbitrary expression displacement of the facial model wireframe fitting, based on detected facial surface feature points. Finally, the facial expression is generated by the vertex displacements of an individual facial model based on estimated muscular contraction parameters. Experimental results reveal that our approach can generate various facial expressions of the individual facial model, which corresponds to the facial expression of the actual face image. Additionally, we can generate the facial expressions of another person by using the muscular contraction parameters of that person in an individual facial model. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(12): 66–75, 2007; Published online in Wiley InterScience (). DOI 10.1002-scj.20647
TL;DR: It is shown that CAN2 based on the proposed algorithm has excellent function approximation performance, and the results are compared to the results of experiments by BPN (Back-Propagation Net), RBFN (Radial Basis Function Net), and SVR (Support Vector Regression), and the effectiveness of the proposed algorithms is verified.
Abstract: The competitive associative net CAN2 is a neural net that provides piecewise-linear approximation through incremental learning of a nonlinear function by the associative net function. The effectiveness of this net has already been shown by applications to function approximation, control, precipitation estimation, and other problems. The learning algorithm essentially includes competitive learning based on the gradient method and suffers from the local solution problem. The purpose of this paper is to circumvent this problem. First, the asymptotic optimality condition, that is, the optimality condition for minimizing the mean-square error of the approximation function, is derived for the case in which the net is composed of a very large number of units. This condition can be used to decide whether the weight assignment each time is close to the optimal solution, and also to suggest weight assignments closer to the optimal solution. The condition is incorporated into a learning algorithm based on the gradient method, resulting in the following learning algorithm: when it is decided that the weight assignment obtained by the gradient method is not close to the optimal solution, the weights of some units are reinitialized so that the weight assignment is closer to the optimal solution. Finally, numerical experiments are performed in which the proposed learning algorithm is applied to several benchmark functions. The effectiveness of the proposed algorithm is verified. The results are compared to the results of experiments by BPN (Back-Propagation Net), RBFN (Radial Basis Function Net), and SVR (Support Vector Regression), and it is shown that CAN2 based on the proposed algorithm has excellent function approximation performance. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(9): 85–96, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10538
TL;DR: This paper has confirmed that the circuits synthesized with the proposed method achieve high performance on gate level simulation and low power consumption, and that the prediction model predicts the sizes of the synthesized circuits accurately enough for practical use.
Abstract: Genetic Algorithms (GA) can be used for various applications including complex computations such as combinatory optimization problems. Such GA applications are desired to be available to information appliances with poor resources by implementing them on dedicated hardware chips like FPGA. In this paper, we propose a method to efficiently design and implement GA applications on FPGA. Our method consists mainly of a parallel and pipelined architecture suitable for various GA applications and a model to predict the size of the synthesized hardware circuits from various parameter values such as the size of the problem and the number of parallel pipelines. In order to facilitate hardware design, we have implemented two tools. The first tool uses our prediction model and calculates parameter values with which the hardware circuits can be synthesized on a specified FPGA device. The second tool generates the RT level VHDL description when the parameter values are given. In order to show efficiency of the proposed method, we have applied our method to the Knapsack Problem and Traveling Salesman Problem. As a result, we have confirmed that the circuits synthesized with our tools achieve high performance on gate level simulation and low power consumption, and that our prediction model predicts the sizes of the synthesized circuits accurately enough for practical use. ©2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(13): 28–38, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20779
TL;DR: A fingerprint verification method which can achieve minutiae matching with highly accurate compensation for rotation and position deviation, as well as high speed and stability is proposed.
Abstract: Fingerprint verification technique is raising high expectations as a means of personal identification based on physical features. This paper proposes a fingerprint verification method which can achieve minutiae matching with highly accurate compensation for rotation and position deviation, as well as high speed and stability. The process flow after binarization of the gray-scale fingerprint image is as follows. (1) Using the fingerprint ridge direction distribution in each local region, highly accurate compensation for the rotation and the position deviation between the enrolled fingerprint and the input fingerprint is performed. (2) The optimal correspondence between enrolled and input minutiae is established by fast combinatorial search. (3) By applying threshold processing to the distances of the fingerprint ridge direction distributions and the minutiae matching rate, the acceptancesrejection of the person is decided. Fingerprint image data for 80 persons×4 fingers×10 samples were collected using a commercial electrostatic capacitance semiconductor fingerprint sensor. The false reject rate and the false acceptance rate are investigated in fingerprint verification experiments, and the effectiveness of the proposed method is demonstrated. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(3): 72–82, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10522
TL;DR: This paper proposes motion compensation using sliced blocks as a new method for macroblock-based motion compensation, where macroblocks are sliced by arbitrary line segments for better agreement in shape between partitions and moving objects, as well as better motion compensation performance.
Abstract: In international standards for video coding such as MPEG-4 and H.264, motion compensation is performed by partitioning macroblocks (square pixel sets) into multiple areas. In this paper, we propose motion compensation using sliced blocks as a new method for macroblock-based motion compensation. The partition shape is not restricted to squares or other patterns; instead, macroblocks are sliced by arbitrary line segments. This produces better agreement in shape between partitions and moving objects, as well as better motion compensation performance. In addition, when partition data are described in bitstream, adaptive prediction of segments based on partition shape of neighbor macroblocks is applied, aiming at reduction of the overhead caused by partition data. In the proposed method, coding is performed at the macroblock level, and therefore, coding mode selection and other related algorithms of conventional rate distortion optimization can be used unaltered. The proposed method was applied to forward-prediction picture coding (basic algorithm of H.264), and an improvement of coding efficiency by about 5p was confirmed. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(7): 12– 22, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20714
TL;DR: An automatic diagram input system which reads the handwritten piping and instrument diagram and recognizes the symbol, character, and line drawn on the diagram and indicates a practically satisfactory performance.
Abstract: This paper describes an automatic diagram input system which reads the handwritten piping and instrument diagram and recognizes the symbol, character, and line drawn on the diagram. The object is a handwritten diagram which is easily understandable by the knowledge of the designer under a very mild constraint. The symbols which are the object of recognition are of more than 200 kinds of various sizes. In addition, a large symbol with unspecified shape may be drawn, and there can be various kinds of lines. To cope with such complex diagrams containing a large amount of handwriting distortions, various diagram processing techniques are proposed. Examples are vector representation with a high accuracy, shaping of various distortions, and the two-stage segmentation, where line, symbol and character are separated hierarchically from the vectorized figure. Other examples are the symbol recognition combining a decision tree and pattern matching, and recognition dictionary based on automatic generation of distorted patterns. The developed diagram input system is a flexible system, in which new symbols can be additionally registered, and complex handwritten diagrams can be recognized. A test was made for the system using the actually written design diagrams. The result indicated a practically satisfactory performance, where the processing time is 3 to 8 min for an A3 sized diagram by a large-scale computer and the recognition rate is above 95 percent.
TL;DR: In the proposed method, the originality of each combination and the novelty perceived by users are estimated from the number of hits on Web pages containing the keywords in question, and the effectiveness of the combinations is judged.
Abstract: Creative activities based on combinations are now being carried on everywhere in the world. Whether or not the creative activity succeeds depends on the combinations used in it. Since there are tremendous numbers of candidate combinations, it is impossible to decide at a glance which should be used. Thus, there must be a method of efficiently discovering combinations which are effective in making creative activity a success. This paper proposes a method of discovering combinations that is effective in supporting creative activity. In the proposed method, the originality of each combination and the novelty perceived by users are estimated from the number of hits on Web pages containing the keywords in question, and the effectiveness of the combinations is judged. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(12): 99–111, 2007; Published online in Wiley InterScience (). DOI 10.1002-scj.20392
TL;DR: A generative text model using Dirichlet Mixtures as a distribution for parameters of a multinomial distribution, whose compound distribution is Polya Mixtures, is proposed and it is shown that the model exhibits high performance in application to statistical language models.
Abstract: We propose a generative text model using Dirichlet Mixtures as a distribution for parameters of a multinomial distribution, whose compound distribution is Polya Mixtures, and show that the model exhibits high performance in application to statistical language models. In this paper, we discuss some methods for estimating parameters of Dirichlet Mixtures and for estimating the expectation values of the a posteriori distribution needed for adaptation, and then compare them with two previous text models. The first conventional model is the Mixture of Unigrams, which is often used for incorporating topics into statistical language models. The second one is LDA (Latent Dirichlet Allocation), a typical generative text model. In an experiment using document probabilities and dynamic adaptation of n-gram models for newspaper articles, we show that the proposed model, in comparison with the two previous models, can achieve a lower perplexity at low mixture numbers. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(12): 76– 85, 2007; Published online in Wiley InterScience (). DOI 10.1002-scj.20629
TL;DR: This paper proposes a license plate recognition method using multivalued template matching, which operates stably even under adverse illumination conditions in which the image contrast is easily degraded, and the recognition rate of the proposed method was much higher than that of the conventional method.
Abstract: This paper proposes a license plate recognition method using multivalued (gray image) template matching, which operates stably even under adverse illumination conditions in which the image contrast is easily degraded. The conventional method has been based on image binarization and the character extraction, which involves the problem that recognition accuracy is greatly lowered when the quality of the input image is degraded. To overcome this problem, the whole configuration of the recognition system is reviewed, and the following two-stage method is considered. First, the individual characters are recognized. Multivalued template matching [log-derivative-matching (LDM) method], including a noise suppression filter based on nonlinear transformation of the gray levels and spatial frequency filtering, is used to scan the whole image. The character candidates and their locations are identified with high accuracy. Second, the license plate is recognized as follows. The likelihood of character existence for each coordinate obtained by image scanning and the character alignment rules in the license plate are compared, and the location with the highest likelihood of being the whole character string is selected. A comparison experiment was performed with more than 10,000 images taken in various outdoor illumination conditions, and it was shown that the recognition rate of the proposed method was much higher than that of the conventional method. The recognition rate for the four main digits was higher than 99p when minimum additional lighting was used, and nearly 95p under very severe illumination conditions. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(3): 49– 61, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20342
TL;DR: The genetic algorithm is applied to the process of deriving a decision algorithm from instances by using rough sets, and a method ofDeriving a simple and useful decision algorithm with a relatively small amount of computation is proposed.
Abstract: With the performance improvement of computers in recent years, the amount of stored available data is rapidly increasing But it is also required that the computer fully utilize the stored resources and perform higher-level intelligent jobs In civil engineering, it is crucial to reuse knowledge which has been accumulated through the experience of engineers, etc For this purpose, it is necessary to establish a method for knowledge acquisition and a method for explicit representation of the acquired knowledge This paper applies the genetic algorithm to the process of deriving a decision algorithm from instances by using rough sets, and proposes a method of deriving a simple and useful decision algorithm with a relatively small amount of computation A decision algorithm is actually derived from the data on accident instances at actual construction sites, and the recognition rate and other performance measures are investigated by the k-fold cross validation method © 2007 Wiley Periodicals, Inc Syst Comp Jpn, 38(4): 39–51, 2007; Published online in Wiley InterScience () DOI 101002sscj20405
TL;DR: This study attempts to extend the lip contour extraction procedure based on the deformable template to the case of a template with left– right asymmetry, and shows that the shape and location of the lip can be extracted precisely and automatically by the proposed method.
Abstract: Studies of facial parts have been conducted in recent years in various fields of medical science and psychology, including studies of the lip contour and location. In fact, in the treatment of cleft lip patients in oral surgery, it is desirable to extract the precise lip contour and to identify its location on the face in order to plan the plastic surgery procedure and to follow up the site after surgery. Aiming at applications to dental treatment, in this study we attempt to extend the lip contour extraction procedure based on the deformable template to the case of a template with left– right asymmetry. An algorithm is also proposed for adjusting the hue and value thresholds in order to improve extraction accuracy and reduce the burden on the user. The proposal is evaluated using sample images of normal subjects and cleft lip patients. It is shown that the shape and location of the lip can be extracted precisely and automatically by the proposed method. In connection with application to dental clinics, we present an example of a quantitative evaluation of the lip shape by using the parameters extracted by the proposed method. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(5): 80– 89, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10668
TL;DR: A new registration algorithm ofgrayscale images and 3D geometric models is proposed by using 2D distance maps that are constructed by using the Fast Marching Method from the boundary of an object on the grayscale image.
Abstract: To create a textured virtual model of a real object, the 3D geometric model that is generally measured by a laser scanner must be aligned with the 2D image that is obtained from image sensors excluding some sensors for which geometric and photometric information are obtained at the same time. One conventional registration method that has been proposed compares the contour lines of the silhouette image of the 3D geometric model and of the 2D image and estimates the relative positions of both according to iterative calculations. However, a method that uses point correspondences of contour points in the 2D image and contour points of the silhouette image must find the nearest neighboring points for each iterative calculation, and the computational cost is high. Therefore, in this paper, the authors propose a new registration algorithm of grayscale images and 3D geometric models by using 2D distance maps that are constructed by using the Fast Marching Method from the boundary of an object on the grayscale image. They also show the effectiveness of the proposed method by performing simulation experiments and experiments using actual images. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(14): 5262, 2007; Published online in Wiley InterScience (). DOI 10.1002/scj.20634
TL;DR: A method that utilizes the properties of vanishing points on lines in photo scenes to join together two images facing two sides front and rear, taken with a fish-eye camera, in order to acquire a full-view spherical image.
Abstract: This paper proposes a method for nonmetric calibration of a fish-eye camera that utilizes the properties of vanishing points on lines in photo scenes to join together two images facing two sides front and rear, taken with a fish-eye camera (for this paper, a camera mounting a fish-eye conversion lens with a field of view extending to more than one hemisphere is called a fish-eye camera), in order to acquire a full-view spherical image. In this method, the geometric properties of vanishing points are first used to obtain initial estimates of the optical center and focal distance on the image, which are the internal parameters for the fish-eye camera. Next, the photo images of the two sides are each mapped onto a spherical surface, and binding conditions, in which true parameter values are used to join the two resulting hemispherical images into a single spherical image, are used to perform fine adjustment of the parameters, based on image correlation of overlapping sections of the front and rear photo images. Experiments using actual still scene images showed the validity of the method, and the proposed method was used to build a system capable of shooting full-view images in real time. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(6): 10–20, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20631 Copyright © 2004 Wiley Periodicals, Inc.
TL;DR: The authors introduced similarity metrics for the similarity between software systems and the Similarity Metrics Measuring Tool (SMMT) for measuring them and applied SMMT to the source code of various UNIX-type operating systems.
Abstract: Given two software systems, it is important to be able to objectively know how much those systems differ. However, it has not been easy to quantitatively measure the mutual differences between two systems. In this paper, the authors introduced similarity metrics for the similarity between software systems and the Similarity Metrics Measuring Tool (SMMT) for measuring them and applied SMMT to the source code of various UNIX-type operating systems. They also performed a cluster analysis from the similarities of the source code of those operating systems to create a dendogram and checked whether the operating systems were classified correctly. The results confirmed that the similarity was an effective index for determining operating system transitions and that the family tree that was obtained represented the classification of the operating systems according to their derivations. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(6): 91–99, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10379
TL;DR: This paper presents a method for real-time 3D human tracking based on the particle filter by incorporating environment models using the multi-viewpoint color and depth images obtained from distributed stereo camera systems and the 3D shape of an indoor environment measured with a range sensor.
Abstract: This paper presents a method for real-time 3D human tracking based on the particle filter by incorporating environment models. We track a human head represented with its 3D position and orientation by integrating the multiple cues from a set of distributed sensors. In particular, the multi-viewpoint color and depth images obtained from distributed stereo camera systems and the 3D shape of an indoor environment measured with a range sensor are used as the cues for 3D human head tracking. The 3D shape of an indoor environment allows us to assume the existing probability of a human head (we call this probability the environment model). While tracking the human head, we consider the environment model to improve the robustness of tracking in addition to the multi-camera's color and depth images. These cues including the environment model are used in the hypothesis evaluation and integrated naturally into the particle filter framework. The effectiveness of our proposed method is verified through experiments in a real environment. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(2): 71–80, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20612
TL;DR: A local kernel is introduced for good handling of local features in SVM and use summation as the integration method to provide robustness to partial occlusion in face detection methods based on SVM.
Abstract: In this paper, a robust face detection method under partial occlusion is proposed. In recent years, the effectiveness of face detection methods using support vector machines (SVM) has been reported, but in conventional algorithms, one kernel is applied to global features extracted from an image. Global features are easily influenced by partial occlusion, and therefore the conventional algorithms appear not to be robust in the presence of occlusion. Good handling of local features is necessary in order to provide robustness to partial occlusion in face detection methods based on SVM. We introduce a local kernel for good handling of local features in SVM and use summation as the integration method. In the experiment, a comparison was made with SVM based on the conventional global kernel and using face images including occlusions and face images including shadows caused by changes in the direction of the light source. The robustness of the proposed method to occlusion was demonstrated. It was also confirmed that faces could be detected from face images including practical occlusions such as sunglasses or scarves. ©2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(13): 39–48, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20614
TL;DR: An algorithm is presented which improves the sensor correlation method used to classify the source illuminating an imaged scene, and it is shown that color compensation based on the relation between the sensor output and the color temperature is useful in such an application.
Abstract: In many fields, including computer vision and image understanding, it is important to obtain knowledge about the illuminating light source from color images. This paper presents an algorithm which improves the sensor correlation method used to classify the source illuminating an imaged scene, and also presents an application. First, we wish to derive a perceptually uniform illumination classification as an alternative to the conventional physical light source classification based on the color temperature scale. For this purpose, the inverse (reciprocal) color temperature scale called mired is used. Second, we calculate the correlation, not between the pixels and the source color region (illuminant gamut), but between the image gamut and the illuminant gamut. Third, a scaling operation for the image is introduced so that the illuminant gamut can be determined by adjusting the intensity difference between images. The validity of the proposed algorithm is evaluated in detail by using an actual image database. Then the proposed source classification method is applied to color compensation. When the color temperature of the source is estimated, the image obtained when the same scene is illuminated by a different source can be predicted. It is shown that color compensation based on the relation between the sensor output and the color temperature is useful in such an application. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(8): 95–108, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10372
TL;DR: The proposed algorithm reduces the number of observed pixel value estimations from the high-resolution image, using an average of pixel values in a divided region, which is about 1.4 to 8.5 times faster than conventional algorithms.
Abstract: A superresolution process produces a high-resolution image from a set of low-resolution images. Reconstruction-based algorithms to produce the high-resolution image which minimizes the difference between observed images and images estimated from the high-resolution image with a camera model have been developed. The reconstruction-based algorithm requires iterative calculation and has a large calculation cost because reconstruction-based superresolution is a large-scale problem. In this paper, a fast algorithm for reconstruction-based superresolution is proposed. The proposed algorithm reduces the number of observed pixel value estimations from the high-resolution image, using an average of pixel values in a divided region. The effect of our proposed algorithm is demonstrated with synthetic images and real images. The results show that the proposed algorithm is about 1.4 to 8.5 times faster than conventional algorithms. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(7): 44–52, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.20662
TL;DR: This paper proposes a calibration method using spherical projection and projection models, which combine the projection function of the fish-eye lens and geometrical distortion models, and can perform calibration quickly and with high accuracy by using patterns of horizontal and vertical parallel lines shown on a liquid crystal display (LCD), which can display such patterns withhigh accuracy.
Abstract: Fish-eye cameras (cameras with a fish-eye lens) have a very wide view angle of about 180° and a great depth of field. Therefore, it is possible to take pictures of broad scenes with a fixed posture, and it is suitable for uses like 3D modeling and surveillance cameras for indoor and outdoor environments. Geometrical and optical calibration of the internal parameters are necessary for improvement of the measuring accuracy in the real world. In this paper, we address the geometrical calibration problem. Some geometrical methods for calibration of the internal camera parameters of fish-eye cameras force the eye camera into a perspective model that is not applicable in wide views of 180° or more. We propose a calibration method using spherical projection and projection models, which combine the projection function of the fish-eye lens and geometrical distortion models. Our method can perform calibration quickly and with high accuracy by using patterns of horizontal and vertical parallel lines shown on a liquid crystal display (LCD), which can display such patterns with high accuracy. We show the effectiveness of our method by an experimental evaluation with a fish-eye camera of orthographic projection type, which obtains high resolution near the edges of the imaging area. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(12): 55–65, 2007; Published online in Wiley InterScience (). DOI 10.1002-scj.20685
TL;DR: A system that automatically measures the cell division pattern of Caenorhabditis elegans embryos from the 1- to 24-cell stages with high levels of objectivity and productivity is presented.
Abstract: Cell division patterns provide crucial information for understanding the mechanisms of development of multicellular organisms. These patterns are measured manually by long-term microscopic observation or by studying large numbers of images taken by using time-lapse microscope systems. Because manual measurement limits the objectivity and productivity of measurement, the comprehensive cell division pattern analysis of gene knockdown embryos has been difficult. Here, we present a system that automatically measures the cell division pattern of Caenorhabditis elegans embryos from the 1- to 24-cell stages with high levels of objectivity and productivity. The system automatically detects the nuclei in a set of four-dimensional differential interference contrast microscope images of a C. elegans embryo by using image processing and measures the cell division pattern by tracking the detected nuclei over time. This system enables a comprehensive cell division pattern analysis of gene knockdown C. elegans embryos. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(11): 12–24, 2007; Published online in Wiley InterScience (). DOI 10.1002-scj.20783