Lin-Cheng Wang

Automatic target recognition using a feature-decomposition and data-decomposition modular neural network

A modular neural network classifier has been applied to the problem of automatic target recognition using forward-looking infrared (FLIR) imagery. The classifier consists of several independently trained neural networks. Each neural network makes a decision based on local features extracted from a specific portion of a target image. The classification decisions of the individual networks are combined to determine the final classification. Experiments show that decomposition of the input features results in performance superior to a fully connected network in terms of both network complexity and probability of classification. Performance of the classifier is further improved by the use of multiresolution features and by the introduction of a higher level neural network on the top of the individual networks, a method known as stacked generalization. In addition to feature decomposition, we implemented a data-decomposition classifier network and demonstrated improved performance. Experimental results are reported on a large set of real FLIR images.

Nonlinear vector prediction using feed-forward neural networks

The performance of a classical linear vector predictor is limited by its ability to exploit only the linear correlation between the blocks. However, a nonlinear predictor exploits the higher order correlations among the neighboring blocks, and can predict edge blocks with increased accuracy. We have investigated several neural network architectures that can be used to implement a nonlinear vector predictor, including the multilayer perceptron (MLP), the functional link (FL) network, and the radial basis function (RBF) network. Our experimental results show that a neural network predictor can predict the blocks containing edges with a higher accuracy than a linear predictor.

A modular neural network vector predictor for predictive image coding

In this paper, we present a modular neural network vector predictor that improves the predictive component of a predictive vector quantization (PVQ) scheme. The proposed vector prediction technique consists of five dedicated predictors (experts), where each expert predictor is optimized for a particular class of input vectors. An input vector is classified into one of five classes, based on its directional variances. One expert predictor is optimized for stationary blocks, and each of the other four expert predictors are optimized to predict horizontal, vertical, 45 degrees , and 135 degrees diagonally oriented edge-blocks, respectively. An integrating unit is then used to select or combine the outputs of the experts in order to form the final output of the modular network. Therefore, no side information is transmitted to the receiver about the selected predictor or the integration of the predictors. Experimental results show that the proposed scheme gives an improvement of 1.7 dB over a single multilayer perceptron (MLP) predictor. Furthermore, if the information about the predictor selection is sent to the receiver, the improvement could be up to 3 dB over a single MLP predictor. The perceptual quality of the predicted images is also significantly improved.

Experimental evaluation of neural, statistical and model-based approaches to FLIR ATR

This paper presents an empirical evaluation of a number of recently developed Automatic Target Recognition algorithms for Forward-Looking InfraRed (FLIR) imagery using a large database of real second-generation FLIR images. The algorithms evaluated are based on convolution neural networks (CNN), principal component analysis (PCA), linear discriminant analysis (LDA), learning vector quantization (LVQ), and modular neural networks (MNN). Two model-based algorithms, using Hausdorff metric based matching and geometric hashing, are also evaluated. A hierarchial pose estimation system using CNN plus either PCA or LDA, developed by the authors, is also evaluated using the same data set.

Composite classifiers for automatic target recognition

Composite classifiers that are constructed by combining a number of component classifiers are designed and evaluated on the problem of automatic target recognition (ATR) using forward-looking IR (FLIR) imagery. Two existing classifiers, one based on learning vector quantization and the other on modular neural networks, are used as the building blocks for our composite classifiers. A number of classifier fu- sion algorithms, which combine the outputs of all the component classi- fiers, and classifier selection algorithms, which use a cascade architec- ture that relies on a subset of the component classifiers, are analyzed. Each composite classifier is implemented and tested on a large data set of real FLIR images. The performance of the proposed composite clas- sifiers are compared based on their classification ability and computa- tional complexity. It is demonstrated that the composite classifier based on a cascade architecture greatly reduces computational complexity with a statistically insignificant decrease in performance in comparison to standard classifier fusion algorithms.

Neural network architectures for vector prediction

A vector predictor is an integral part of the predictive vector quantization (PVQ) scheme. The performance of a predictor deteriorates as the vector dimension (block size) is increased. This makes it necessary to investigate new design techniques in order to design a vector predictor which gives better performance when compared to a conventional vector predictor. This paper investigates several neural network configurations which can be employed in order to design a vector predictor. The first neural network investigated in order to design the vector predictor is the multi-layer perceptron. The problem with multi-layer perceptron is the long convergence time which is undesirable when the on-line training of the neural network is required. Another neural network called functional link neural network has been shown to have fast convergence. The use of this network as a vector predictor is also investigated. The third neural network investigated is a recurrent type neural net. It is similar to the multi-layer perceptron except that a part of the predicted output is fed back to the hidden layer/layers in an attempt to further improve the current prediction. Finally, the use of a radial-basis function (RBF) network is also investigated for designing the vector predictor. The performances of the above mentioned neural network vector predictors are evaluated and compared with that of a conventional linear vector predictor.

Segmentation based wavelet coding of digital images

In this paper, we present a segmentation based wavelet coding scheme, in which an image is segmented into two regions: stationary areas (background) and the areas containing edge information (foreground). These regions are then encoded independently using two dedicated encoders that are optimized for each region. A 2-D edge operator is used for segmenting the image. We use the embedded zerotree wavelet (EZW) algorithm for encoding the background due to its good performance on stationary areas. The foreground area is, however, encoded using a predictive residual vector quantizer (PRVQ). Experimental results show that the proposed technique improves the quality of the reconstructed images, both numerically (in terms of mean square error) and perceptually when compared to EZW at the same bit rate.

Finite-state residual vector quantization

This paper presents a new FSVQ scheme called Finite-State Residual Vector Quantization (FSRVQ) in which each state uses a Residual Vector Quantizer (RVQ) to encode the input vector. Furthermore, a novel tree- structured competitive neural network is proposed to jointly design the next-state and the state-RVQ codebooks for the proposed FSRVQ. Joint optimization of the next-state function and the state-RVQ codebooks eliminates a large number of redundant states in the conventional FSVQ design; consequently, the memory requirements are substantially reduced in the proposed FSRVQ scheme. The proposed FSRVQ can be designed for high bit rates due to its very low memory requirements and low search complexity of the state-RVQs. Simulation results show that the proposed FSRVQ scheme outperforms the conventional FSVQ schemes both in terms of memory requirements and perceptual quality of the reconstructed image. The proposed FSRVQ scheme also outperforms JPEG (current standard for still image compression) at low bit rates.

Rate-constrained modular predictive residual vector quantization

This paper investigates a novel modular image coding paradigm using residual vector quantization (RVQ) with memory that incorporates a modular neural network vector predictor in the feedback loop. A modular neural network predictor consists of several expert networks, where each expert network is optimized for predicting a particular class of data, and an integrating unit that mixes the outputs of the expert networks in order to form the final output of the prediction system. The vector quantizer also has a modular structure. The proposed modular predictive RVQ (MPRVQ) is designed by imposing a constraint on the output rate of the system. Experimental results show that the modular PRVQ outperforms simple PRVQ by as much as 1 dB at low bit rates. Furthermore, for the same PSNR, the modular PRVQ reduces the bit rate by more than a half when compared to the JPEG algorithm.

Automatic target recognition using a modular neural network with directional variance features

A modular neural network classifier has been applied to the problem of automatic target recognition (ATR) using forward- looking infrared (FLIR) imagery. This modular network classifier consists of several neural networks (expert networks) for classification. Each expert network in the modular network classifier receives distinct inputs from features extracted from only a local region of a target, known as a receptive field, and is trained independently from other expert networks. The classification decisions of the individual expert networks are combined to determine the final classification. Our experiments show that this modular network classifier is superior to a fully connected neural network classifier in terms of complexity (number of weights to be learned) and performance (probability of correct classification). The proposed classifier shows a high noise immunity to clutter or target obscuration due to the independence of the individual neural networks in the modular network, Performance of the proposed classifier is further improved by the use of multi-resolution features and by the introduction of a higher level neural network on the top of expert networks, a method known as stacked generalization.