By Topic

Cellular Neural Networks and their Applications, 1996. CNNA-96. Proceedings., 1996 Fourth IEEE International Workshop on

Date 24-26 June 1996

Filter Results

Displaying Results 1 - 25 of 87
  • 1996 Fourth IEEE International Workshop on Cellular Neural Networks and their Applications Proceedings (CNNA-96)

    Save to Project icon | Request Permissions | PDF file iconPDF (450 KB)  
    Freely Available from IEEE
  • Index of authors

    Page(s): 493 - 495
    Save to Project icon | Request Permissions | PDF file iconPDF (84 KB)  
    Freely Available from IEEE
  • Combinatorial topology and qualitative dynamics in cellular neural networks

    Page(s): 191 - 195
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (240 KB)  

    It is shown that the whole CNN invariant manifold structure can be described by the incidence relations between the cells of various orders of the n-dimensional cube and its dual polyhedron. The bifurcations of codimension 1 have a counterpart in transformations of the above geometrical structures due to shrinking or disappearing of specific cells. The geometrical interpretation allows us to predict some qualitative properties of a very general nature for the CNN dynamics View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • SCNN: a universal simulator for cellular neural networks

    Page(s): 255 - 259
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (232 KB)  

    In this paper a universal simulator for cellular neural network (CNN) is presented. CNN with nonlinear and delay-type templates can be simulated precisely with SCNN, practically without any limitations. Furthermore different training algorithms for networks with translation variant and invariant templates are implemented in SCNN. As an example, parameter deviations of a template have been reduced by training. Simulation and training results are discussed in detail View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Morphological operators on the CNN Universal Machine

    Page(s): 151 - 156
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (432 KB)  

    We show that the basic morphological operators can be implemented on the CNN Universal Machine. This includes binary morphological operators that were tested and verified on a working CNN Universal chip. We also show different implementation methods for grayscale morphology using different template types View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CNN chip set architectures and the visual mouse

    Page(s): 487 - 492
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (308 KB)  

    The cellular neural network (CNN) Universal Machine architecture when implemented in VLSI chips needs special interfaces to provide for efficient system performance both in time (speed) and space (image size). The CNN chip set architectures described solve this problem, including interfacing the possibly analog input sensors and digital output and control. Various forms of CNN Engines are presented embedding CNN chip sets. A new device, the Visual Mouse, a hand held visual supercomputer, is also presented which exploits the genuine features of CNN Engines View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Large-image CNN hardware processing using a time multiplexing scheme

    Page(s): 405 - 410
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (360 KB)  

    The state of the art work in cellular neural networks (CNN) has concentrated on VLSI implementations without really addressing the “systems level”. While efficient implementations have been reported, no reports have been presented on the use of these implementations for processing large complex images. The work hereby presented introduces a strategy to process large images using small CNN arrays. The approach, time-multiplexing, is prompted by the need to simulate hardware models and test hardware implementations of CNN. For practical size applications, due to hardware limitations, it is impossible to have a one-on-one mapping between the CNN hardware processors and all the pixels in the image involved. This paper presents a practical solution by processing the input image block by block, with the number of pixels in a block being the same as the number of CNN processors in the hardware. Image processing results obtained from an actual IC test-chip prototype using this scheme are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A learning algorithm for the dynamics of CNN with nonlinear templates. I. Discrete-time case

    Page(s): 461 - 466
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (196 KB)  

    A learning algorithm for the dynamics of discrete-time cellular neural networks (DTCNN) with gradient-based nonlinear templates is presented. For modeling the dynamics of nonlinear spatio-temporal systems with DTCNN, the algorithm is applied to find the network parameters. Results for two different nonlinear discrete-time systems are discussed in detail View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A 9×9 multi-chip board for cellular neural networks

    Page(s): 261 - 266
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (408 KB)  

    Cellular neural networks represent one of the most interesting neural network class for VLSI implementations. Their main feature is the local interconnections among cells that is really appealing for hardware realizations. This paper presents a multichip 9×9 CNN board which is made up of nine 3×3 DPCNN chips. This 9×9 CNN board is fully controllable by a Personal Computer. A proper software program is presented which enables the selection of all the entries of the templates, the input and initial voltage, and the steady-state voltage acquisitions View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Templates design for high quality digital images by discrete time cellular neural network

    Page(s): 333 - 338
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (256 KB)  

    This paper describes reduction methods for false-outlines, overload edge noise and flicker based on optimal templates design including motion detection by multi-level discrete time cellular neural networks (DTCNN). The templates are designed by the minimization of Lyapunov-energy function to minimize the least-square distortion for digital still and moving images. High quality of the digital images was demonstrated by using a Canon ferroelectric liquid crystal display (FLCD) which is the first commercial display device with bilevel for red, green, blue or white View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CNNs with radial basis input function

    Page(s): 231 - 236
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (320 KB)  

    This paper proposes a cellular neural network (CNN) model with radial basis input function (radial basis input CNN) for improving function approximation ability of CNNs. The model can be viewed as a cascade of two units: the first unit is a multi-input, multi-output radial basis function network (RBFN), the second unit is the original CNN model. The weights and centers of the RBFN unit are chosen identical for all RBFN outputs yielding a space-invariant connection weight pattern over the network. With such a weight sharing property, the proposed model becomes a special kind of nonlinear B-template CNN. The ability of the radial basis input CNN model in approximation to functions as its input-(steady state) output mapping is examined on an edge detection task for noisy images. A modified version of the recurrent perceptron learning algorithm (RPLA) is used for the training radial basis input CNN View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Structure verification on photolithographic masks based on tolerance criteria by cellular neural networks

    Page(s): 145 - 149
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (272 KB)  

    First of all, the problem of structure verification is introduced with respect to tolerance criteria and their verifications. After that, the basic principle of a verification method for cellular neural networks is motivated by means of local restrictions. Then their solution is presented on the basis of local operators which are only designed with the help of local restrictions of the design rules of the mask structures and their tolerance zones. After that, the result of the successful use of the method is demonstrated on manufacturing and calibration masks. Finally, the debate discusses the competitive position of the method with reference to two standard methods. Last of all, the summary indicates future improvements View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Global stability of CNNs with delay

    Page(s): 187 - 190
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (168 KB)  

    We give a new sufficient condition for the global stability of the equilibrium point for cellular neural networks with delay (DCNNs). The significance of this condition is that the existence, uniqueness and stability of the equilibrium point is not affected by the delay parameter. This condition is also weaker than some previous global stability results View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The edge regularization of noised image method

    Page(s): 207 - 211
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (228 KB)  

    A model of extraction and regularization of edges of noisy images is presented. Edge is defined as a local maximum of the image gradient. The model is based on a formation of local dependences between pixels of the image, so called pattern and step dynamic approximation of the value of the brightness of individual pixels. The dynamics of the model are defined by an adequately constructed function of cost (further on called energy, due to an analogy with potential energy). Energy is specified from the use of the sum of adversities of scalar products of the neighboring vectors of the gradient. This process can be easily implemented in a cellular neuron network of a suitable design. Regularization of the image is an important phase of image processing among others image recognition View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On globally asymptotically stable continuous-time CNNs for adaptive smoothing of multidimensional signals

    Page(s): 351 - 356
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (460 KB)  

    We present a theoretical framework from which an approach to nonlinear, locally-adaptive smoothing of multi-dimensional signals has been derived which exhibits properties favourable to any application: unique solution, data adaption, presentation of signal structure, continuous dependency of the result on both the input signal and few parameters, and effective control of parameters. We also show that i) the FEM discretisation nicely inherits the properties of the continuous notation, and that ii) the discretised version represents a globally asymptotically stable network. We then explicate the embedding of our approach in the continuous-time CNN paradigm of Roska and Chua (1992) and provide results from our simulations. Lastly we report on ongoing work towards CNN circuit design such as to render possible real-time processing in the future-a desideratum in computer vision system design View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CNN-based retinal model uncovers a new form of edge enhancement in biological visual processing

    Page(s): 303 - 307
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (500 KB)  

    Visual processing in the retina is mediated by complex interactions between millions of neurons that shape the visual message in both space and time. Conventional neurobiological method study single cells, and activity is recorded in response to simple stimuli as a function of time. By modeling retina activity using cellular neural network (CNN) one can begin to think about retinal interactions in space/time, and consider the activity of large populations of cells as the deformation of surfaces. Using CNN we predicted a powerful form of edge enhancement mediated by a novel space-time interaction. This is the first known form of edge enhancement that is not mediated by lateral inhibition. So far we have examined this mechanism in salamander retina, but hope to extend these results to mammalian retinas as well View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CNN implementation of seed growth algorithm for fuzzy segmentation of images

    Page(s): 197 - 200
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (276 KB)  

    This paper describes a cellular neural network implementation of the seed growth algorithms for fuzzy segmentation of images. Some examples with medical images are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improved scoring and semi-automatic screening of human peripheral blood chromosomes by CNN visual system

    Page(s): 99 - 102
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (732 KB)  

    Many of the microscopic image processing tasks can be well implemented in the cellular neural net Universal Machine (CNN UM) architecture. We have developed a complex system for chromosome analysis. Our method, when implemented in hardware containing VLSI chips, can execute some important image processing steps at superior speed. The recent simulator based system is capable of helping the reliable chromosome analysis View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Application of reversible discrete-time cellular neural networks to image copyright labelling

    Page(s): 19 - 24
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (308 KB)  

    In this paper, we proposed a discrete-time cellular neural network (DTCNN) structure for the labelling of digital images. First, we present the concept and the structure of reversible DTCNN, which can be used to generate 2D binary random image sequences. Then both the original image and the copyright label, which is often another binary image, are used to generate a binary random key image. The key image is then used to scramble the original image. Due to the reversibility of a reversible DTCNN, the same reversible DTCNN is used to recover the copyright label from a labelled image. Due to the high speed of a DTCNN chip, our method can be used to label image sequences, e.g., video sequences, in real time. Computer simulation results are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Information processing with lossless cellular neural networks

    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (44 KB)  

    Summary form only given. The eigenvalues of a symmetric tridiagonal matrix can be computed via an iterative diagonalization with the aid of the QR-algorithm. Interpolating the matrices of subsequent iteration steps with continuous (time) trajectories leads to the concept of matrix flows on manifolds. This can be viewed as the transients of a nonlinear dynamical system. In the fixed point of the dynamical system, the desired eigenvalues are found. Mapping the nonlinear ODE's onto physical structures, lossless nonlinear dynamical circuits are obtained. These circuits can be interpreted as CNN's with nonlinear templates. While these lossless CNNs are interesting by themselves, the important question arises, whether a robust implementation without any power supply and, therefore, without dissipation is practically feasible indeed. Conventional realizations would necessitate (nonlinear) inductors and would not be well suited for solid state silicon implementations. Any simulation of the inductors with the aid of transistors will not alleviate the problem, because the resulting circuits will not be truly lossless any more. Future nanoscale quantum devices open up new possibilities of low loss nanoelectronic computing structures View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • ACE: a digital floating point CNN emulator engine

    Page(s): 273 - 278
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (424 KB)  

    The architecture of ACE, a multiprocessor analogic cellular neural network (CNN) emulator engine consisting of 2 to 16 TMS320C40 floating point DSPs is introduced. The engine containing up to 512 Mbyte RAM (enough to store a 512×512×512 sized CNN cube) which can be controlled through its SCSI port. It can either accelerate the multilayer CNN simulator CNNM or be accessed directly from the high level, C-based analogic CNN language ACL to achieve the simulation speed of ~2.8 μsec/cell/iteration/DSP for 3×3 linear templates View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Data parallel software simulation of cellular neural networks

    Page(s): 267 - 271
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (288 KB)  

    A data parallel model for the parallel simulation of cellular neural networks is proposed. This approach is based on the SPMD (single-program multiple data) model and utilizes the highly specialized programming environment of the Vienna FORTRAN compilation system (VFCS) for the parallel process. It allows an easy development of the simulation system as a sequential program attributed with data distribution information shifting the difficult parallel process to the compiler. An analysis of the run-time behavior of the developed parallel programs for the simulation of cellular neural networks for different data distribution schemes is given, which justifies the efficiency of the proposed approach View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A fully optically addressable connected component detector in CMOS

    Page(s): 439 - 443
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (196 KB)  

    A connected component detector cellular neural network array chip with full optical input and output is described. The chip is based on a hybrid integration of gallium arsenide quantum well modulators on CMOS silicon circuitry. The circuit design and simulation results are presented. An optical demonstration system is also described View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On a novel adaptive self organizing network

    Page(s): 41 - 46
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (212 KB)  

    In this paper a new algorithm is presented in order to overcome the stability vs. formation ability dilemma of competitive learning. This algorithm is based on growing cell structures of self-organizing mapping. The new algorithm is effective for endless learning and automatic classification. Applying the algorithm in the case where the input pattern is changed temporally, we have confirmed that it has much better performance than conventional algorithms View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Optimizing the morphological design of discrete-time cellular neural networks

    Page(s): 339 - 343
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (276 KB)  

    The morphological design of discrete-time cellular neural networks (DTCNNs) has been presented in a companion paper (1996). DTCNN templates have been given for the elemental morphological operators. One way to obtain realizations for more complex operators is cascading the DTCNN equivalences of the constituent elemental operators. Here it is shown that this straightforward mapping mostly yields a non-optimal solution with respect to the required amount of hardware. A hardware reduction scheme of morphologically designed DTCNNs is proposed which includes the introduction of time variant templates and the identification of non-elementary expressions for which a single layer DTCNN exists View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.