Search

A statistical approach for modeling the code vectors designed using a supervised learning neural network is proposed in this paper. Since wavelet-based compression is more robust under transmission and decoding errors, the proposed work is implemented in the wavelet domain. Two crucial issues in compression methods are the coding efficiency and the psycho visual quality achieved while modeling different image regions. In this paper, a high performance wavelet coder which provides a new framework for handling these issues in a simple and effective manner is proposed. First the input image is subjected to wavelet transform. Then the transformed coefficients are subjected to Quantization followed by the well known Huffman Encoder. In the Quantization process, initially a codebook is designed using Learning Vector Quantizer. Since codebook is an essential component for the reconstructed image quality and also to exploit the spatial energy compaction of the codevectors, the codebook is further modeled using Savitzky–Golay polynomial. Experimental results show that the proposed work gives better results in terms of PSNR that are competitive with the state-of-art coders in literature. [ABSTRACT FROM AUTHOR]

In this paper, a novel unsupervised competitive learning algorithm, called the centroid neural network adaptive resonance theory (CNN-ART) algorithm, is proposed to relieve the dependence on the initial codewords of the codebook in contrast to the conventional algorithms with vector quantization in lossy image compression. The design of the CNN-ART algorithm is mainly based on the adaptive resonance theory structure, and then a gradient-descent-based learning rule is derived so that the CNN-ART algorithm does not require a predetermined schedule for learning rate. Furthermore, the appropriate initial weights obtained by the CNN-ART algorithm can be applied as an initial codebook for the Linde–Buzo–Gray (LBG) algorithm such that the compression performance can be greatly improved. In this paper, the extensive simulations demonstrate that the CNN-ART algorithm does outperform other algorithms like LBG, self-organizing feature map and differential competitive learning. [Copyright &y& Elsevier]

Abstract: The disadvantage of the generalized learning vector quantization (GLVQ) and fuzzy generalization learning vector quantization (FGLVQ) algorithms is discussed in this paper. And a revised generalized learning vector quantization (RGLVQ) algorithm is proposed to overcome the disadvantage of GLVQ and FGLVQ. Furthermore, by introducing a stimulating coefficient in completing step, a new competing technique to improve the performance of the LVQ neural network is proposed also. The proposed algorithms are tested and evaluated using the IRIS data set. And the efficiency of the proposed algorithms is also illustrated by their use in codebook design for image compression based on vector quantization, and the training time for RGLVQ algorithm is reduced by 10% as compared with FGLVQ while the performance is similar. The new competing technique is also used to generate codebook and PSNR is improved in experiments. [Copyright &y& Elsevier]

Convolutional neural networks (CNN) are the state-of-the-art method for action recognition in various kinds of datasets. However, most existing CNN models are based on lower-level handcrafted features from gray or RGB image sequences from small datasets, which are incapable of being generalized for application to various realistic scenarios. Therefore, we propose a new deep learning network for action recognition that integrates quaternion spatial-temporal convolutional neural network (QST-CNN) and Long Short-Term Memory network (LSTM), called QST-CNN-LSTM. Unlike a traditional CNN, the input for a QST-CNN utilizes a quaternion expression for an RGB image, and the values of the red, green, and blue channels are considered simultaneously as a whole in a spatial convolutional layer, avoiding the loss of spatial features. Because the raw images in video datasets are large and have background redundancy, we pre-extract key motion regions from RGB videos using an improved codebook algorithm. Furthermore, the QST-CNN is combined with LSTM for capturing the dependencies between different video clips. Experiments demonstrate that QST-CNN-LSTM is effective for improving recognition rates in the Weizmann, UCF sports, and UCF11 datasets. [ABSTRACT FROM AUTHOR]