<title>Adaptive image segmentation by quantization</title>

We have developed a self-creating and organizing unsupervised learning algorithm for artificial neural networks. In this study, we introduce SCONN and SCONN2 as two versions of self-creating and organizing neural network (SCONN) algorithms. SCONN creates an adaptive uniform vector quantizer (VQ), whereas SCONN2 creates an adaptive nonuniform VQ by neural-like architecture. SCONN's begin with only one output node, which has a sufficiently wide activation level, and the activation level decrease depending upon the time or the activation history. SCONN's decide automatically whether to adapt the weights of existing nodes or to create a new "son node." They are compared with two famous algorithms-the Kohonen's self organizing feature map (SOFM) (1988) as a neural VQ and the Linde-Buzo-Gray (LBG) algorithm (1980) as a traditional VQ. The results show that SCONN's have significant benefits over other algorithms.

Kohonen's Self Organizing Feature Maps with variable learning rate. Application to image compression

An investigation of local energy surface detection integrated with neural network techniques for image segmentation is presented, as applied in the feature extraction of chromosomes from image datasets obtained using an experimental confocal microscope. Use of the confocal microscope enables biologists to observe dividing cells (living or preserved) within a three-dimensional (3-D) volume, that can be visualised from multiple aspects, allowing for increased structural insight. The Nomarski differential interference contrast mode used for imaging translucent specimens, such as chromosomes, produces images not suitable for volume rendering. Segmentation of the chromosomes from this data is, thus, necessary. A neural network based on competitive learning, known as Kohonen's self-organizing feature map (SOFM) was used to perform segmentation, using a collection of statistics or features defining the image. Our past investigation showed that standard features such as the localized mean and variance of pixel intensities provided reasonable extraction of objects such as mitotic chromosomes, but surface detail was only moderately resolved. In this current work, a biologically inspired feature known as local energy is investigated as an alternative image statistic based on phase congruency in the image. This, along with different combinations of other image statistics, is applied in a SOFM, producing 3-D images exhibiting vast improvement in the level of detail and clearly isolating the chromosomes from the background. Index Terms-DIC, differential interference contrast, feature extraction, feature space, image segmentation, local energy, Morlet wavelet, phase congruency, self organizing feature map, SOFM.

We examine the performance of several neural network vector quantization (VQ) methods on image coding. The VQ methods we look at are Kohonen's Self-Organizing Feature Map (KSOFM), Frequency-Sensitive Competitive Learning (FSCL), and Self-Creating and Organizing Neural Network (SCONN). We also look at variations of these algorithms by combining different methods. Our simulation results show that the best performance is achieved by the SCONN and the combination KSOFM and FSCL.

A new predictive vector quantization (PVQ) technique capable of exploring the nonlinear dependencies in addition to the linear dependencies that exist between adjacent blocks (vectors) of pixels is introduced. The two components of the PVQ scheme, the vector predictor and the vector quantizer, are implemented by two different classes of neural networks. A multilayer perceptron is used for the predictive cornponent and Kohonen self-organizing feature maps are used to design the codebook for the vector quantizer. The multilayer perceptron uses the nonlinearity condition associated with its processing units to perform a nonlinear vector prediction. The second component of the PVQ scheme vector quantizes the residual vector that is formed by subtracting the output of the perceptron from the original input vector. The joint-optimization task of designing the two components of the PVQ scheme is also achieved. Simulation results are presented for still images with high visual quality.

One of the prerequisites of effective operation of multiphase flow systems is the need to identify flow regimes. A number of flow regime classification models have been reported in the literature based on the subjective and ambiguous interpretation of visual observations. With the advent of microprocessors and recent advances in signal processing, pattern recognition and neural network techniques, we are now on the threshold of finding more objective and quantitative technique for flow regime identification and flowrate measurement in multiphase systems. This paper presents a technique for the classification of flow regimes in horizontal air-water two-phase flow lines by the application of the Kohonen self-organising feature map (KSOFM).The principle of the technique rests on the characterisation and classification of turbulent pressure signals in relation with flow regimes. Characterisation means deriving stochastic features from time domain signals which are likely to be associated with specific physico-fluid dynamic groups based on space-time average values. Classification means defining the boundaries of the fluid regimes as represented by the fluid dynamic groups.KSOFM is known to be capable of organising patterns from an arbitrary n-dimension space into lower dimension space and preserving the topological relationships among the features. In this case, a KSOFM with 8*8 neurons was selected and trained with stochastic features derived from the absolute pressure signals obtained from a range of flow regimes. The resulting feature map succeeded in organising samples from similar flow regimes into same or adjacent neurons and separating samples from different flow regimes into different areas.The samples used in the tests were obtained in laboratory measurements conducted in a horizontal 2 inch diameter air-water flow line. KSOFM identified the flow regimes observed visually as bubbly, slug and wavy/stratified flow regimes as well as the more elusive transition regimes. For future work, it is proposed that the KSOFM offers a preprocessing capability in separating training regions into sub-regions where within each sub-region other supervised network models, such as the back propagation neural networks can be trained to identify primary quantities of interest such as average mass/volumetric flow rates of individual phases.

Kohonen maps are self-organizing neural networks that categorize input data, capturing its topology and probability distribution. Efficient hardware implementations of such maps require the definition of a certain number of simplifications to the original algorithm. In particular, multiplications have to be avoided by means of choices in the distance metric, the neighborhood function and the set of learning parameter values. In this paper, one-dimensional and bi-dimensional Kohonen maps with exponential neighboring function and Cityblock and Chessboard norms are defined, and their hardware architecture is presented. VHDL simulations and synthesis on an FPGA of the proposed architectures demonstrate both satisfactory functionality and feasibility

Perception in natural systems is a highly active process. In this paper, we adopt the strategy of natural systems to explore objects for 3D object recognition using robots. The exploration of objects enables the system to learn objects from different viewpoints, which is essential for 3D object recognition. Exploration furthermore simplifies the segmentation of the object from its background, which is important for object learning in real-world environments, which are usually highly cluttered. We use the scale invariant feature transform (SIFT) as the basis for our object recognition system. We discuss our active vision approach to learn and recognize 3D objects in cluttered and uncontrolled environments. Furthermore, we propose a model to reduce the number of SIFT keypoints stored in the object database. It is a known drawback of SIFT that the computational complexity of the algorithm increases rapidly with the number of keypoints. We discuss the use of a growing-when-required (GWR) network, which is based on the Kohonen self organizing feature map, for efficient clustering of the keypoints. The results show successful learning of 3D objects in a cluttered and uncontrolled environment. Moreover, the GWR-network strongly reduces the number of keypoints.

The authors developed a new method in order to automatically segment magnetic resonance images (MRIs) of the head. The main tissues, such as scalp, brain and skull, are recognised. The method is based on a Kohonen self organising feature map which performs a cluster of the image areas into three main classes. The network, after being trained, is successfully operated on the test set. The network performances do not depend on the MRI apparatus producing the images set. The network classes are properly matched and processed in order to obtain slices containing the desired tissues. The proposed method has been developed in the frame of a project for the 3-dimensional reconstruction of selected surfaces.

The paper presents applications of neural network algorithms to the design of an adaptive vector quantizer. Vector quantization has been applied to the problem of displaying natural images with a reduced set of colors (colormap) and to the interframe coding of image sequences. The first step was to test classical Linde Buzo Gray (LGB), self organizing feature maps (SOFM) and Competitive Learning (CL) algorithms for the codebook design. The best results for the reconstructed quality image and the computational time are obtained using a CL algorithm with a new initialization strategy that solves the problem of underutilized nodes. An adaptive vector quantization algorithm is proposed and tested in a motion compensated image coder. The results of the simulations are very promising. In fact the coder performance, compared with that using a fixed VQ, is considerably improved and the subjective quality of the coded images is much better than that obtained using standard vector quantization, especially when rapid motion is present in the scene.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Image data compression is essential for a number of applications that involve transmission and storage. One technique that has been recently extensively investigated is vector quantization (VQ). One class of neural networks (NN) structures, namely competitive learning networks appears to be particularly suited for VQ. One main feature that characterizes NN training algorithms is that the VQ codewords are obtained in an adaptive manner. In this paper, a new competitive learning (CL) algorithm called the Threshold Competitive Learning (TCL) is introduced. The algorithm uses a threshold to determine the codewords to be updated after the presentation of each input vector. The threshold can be made variable as the training proceeds and more than one threshold can be used. The new algorithm can be easily combined with other NN training algorithms such as the Frequency-Sensitive competitive learning (FSCL) or the Kohonen Self-Organizing Feature Maps (KSFM). The new algorithm is shown to be efficient and yields results comparable to the famous traditional LBG algorithm.

A self-organizing neural network model is proposed for pattern classification for any given data sets without a priori information about the number of clusters or cluster centers. This system is based on extensions of Kohonen's self-organizing feature map model. It adjusts weights between the input layer and the distance layer until it is stabilized by utilizing feedback information for specific neurons. Then it classifies each input according to the distance between the weights and the normalized input using J.C. Bezdek's (1981) membership value equation. The final weights between the input layer and the distance layer become the coordinates of the cluster centers. Some experimental studies are presented for a sample data set and a color image. >

In this paper, a new method of vector quantizer design for image compression using Generic codebook and wavelet transformation is proposed. In the proposed method, Self Organizing Feature Map (SOFM) is used for initial codebook generation. A new scheme of wavelet transformation based Vector Quantization (VQ) technique is proposed to replace the SOFM code vectors by VQ code vectors. The proposed wavelet transform is used to generate wavelet coefficients which are then converted into VQ code vectors. Discrete Cosine Transformation based vector quantization technique is proposed in the existing image compression algorithms with low quality images with greater amount of information loss. Hence to increase the psycho visual quality of the reconstructed image wavelet transformation based vector quantization technique is proposed in this paper. Performance of the proposed work is tested with varying codebook size and various training images. Experimental results show that the reconstructed images obtained by the proposed method are of good quality with better compression ratio and higher Peak Signal–to–Noise Ratio.

Conventional techniques used for identification and equalization of nonlinear M-ary PSK digital satellite channels are based on linear or nonlinear filtering devices (e.g. tapped delay line equalizers, Volterra series approaches). This paper uses a new technique based on the vector neural network (VNN) and Kohonen (1989) self organizing feature map. We have used a VNN for adaptive equalization and identification of the satellite channel. The decision process is performed by a Kohonen map.

A new Lagrangian formulation with entropy and codebook size was proposed to extend the Lagrangian formulation of variable-rate vector quantization. We use the new Lagrangian formulation to perform clustering and to find the number of clusters by fitting mixture models to data using vector quantization. Experimental results show that the entropy and memory constrained vector quantization outperforms the state-of-the-art model selection algorithms in the examples considered.