The high frequency part of wavelet transform reflects the detail of edges. And median filter is a kind of nonlinear stack filter, it has good abilities in keeping edge and filtering clutter. Based on the analysis of the characteristics of the images of Synthetic Aperture Radar (SAR), this paper puts forward a combined method for edge s detection of SAR images. This method is composed of wavelet transform, fuzzy median filter, adaptive thresholding and differential gradient operator. In using fuzzy median filter, this paper gives out a new method for adaptive selecting the factors of fuzzy median filter. The experimental results show the effectiveness and stability of this method.

In this paper, the question of the parameter choice of RDM (reflectivity displacement method) for the radar system that has a narrow azimuth antenna beam is discussed and the important factors that determine the choice of the parameters are proposed. Experiments using raw data of L SAR that has an azimuth beam of only 9° show that good detection result can be achieved when applied these principles for parameter choice.

Base on the analysis of SAR linear FM waveform and Doppler history, a modified spaceborne SAR quicklook algorithm is proposed, the algorithm s mathematical model is also constructed. Then a method of how to mosaic the sub-image together is given by analyzing Doppler history. Finally, by processing ERS-1 raw data, the results indicates that the quicklook algorithm is precise and efficient.

In order to save the computational time of the image correlation matching, genetic algorithm is imported. Both binary coding and natural number coding genetic algorithms are implemented for correlation matching respectively. The results show that natural number-coding method not only is practical, but also runs faster than binary coding method in the application of image correlation matching.

Based on digital image processing and analysis, and computer graphics, a task of rebuilding and exhibiting unearthed relics models in three dimensions has been studied by vising 3D modeling software and 2D digital image processing software in this paper. With the developed procedure for computer-rebuilding relics in three dimensions, rebuilding a cultural relic (two-sheep vessel) is shown in this paper. It is proved that this technique can be extended to rebuild other cultural relics.

The new video coding standard MPEG4 is enabling content-based functionality. It decomposes sequences into video object planes (VOPs) so that each VOP represents one moving object. This paper presents a new automatic video sequence segmentation algorithm that extracts moving objects. The core of this algorithm is an object tracker that matches a two-dimensional binary model of the object against subsequent frames using the Hausdorff distance. The initial model is derived automatically and a new model update method based on the concept of the moving connected components is proposed. The stationary background is removed by a filtering technique. Finally, the VOPs are extracted from the binary model sequence.

In this paper, a modified scheme for small memories and low-power image and video coding is presented. It is based on vector quantization. In 1998, K. Masselos uses a small codebook, and by using simple but efficient transformations to the codewords during coding procesa the small codebook is extended, thus compensated for the quality degradation intro- duced by the small codebook size. In this way, the power consumption is reduced. Compared to the scheme proposed by K. Masselos, we store the quantized DCT coefficients of codewords in the codebook instead of the codewords itself, and the distortion calculation is performed in the DCT field respectively. Because we use only a part of coefficients to calculate the distances, the distortion calculation is simplified. Furthermore, we modified the way of designing the small (odebook; the possible redundancy in K. Masselos’s codebook is eliminated.

This paper analyses various objective measures for the prediction of subjective quality of speech. A new Perception-based Bark Spectral Distortion (PBSD) parameter is pre- sented, which takes the masking property of human ear into consideration, to predict Mean Opinion Score(MOS) of speech quality. Experiments prove that this map from objective mea-sure to subjective MOS based on the calculation of PBSD has rather small prediction error. The PBSD parameter is applied to designing new speech codec in place of MSE parameter and the subjective quality of decoded speeches is improved.

The eigenvector of n-cube graph is found and then a kind of Hopfield networks with n-cube as its structure and connection situation, called n-cube network, are analysed theoretically for the attractions states/cycles and their distribution in the state space of the network. The analysis shows that the eigenvectors of the connection matrix of the network and their linkages are, in general, either the attraction states or attraction cycles of the net, and the distribution of these states/cycles is symmetric and uniform in state space of the network.

In this paper, a neural network with strong kernel and continuously time delay is mainly investigated. A switch from stability to instability may occur for certain range of the parameters arid must then be followed by a switch back to stability. Bifurcation phenomena of this model are also investigated. Using the mean time delays as a bifurcation parameter, authors have proven that Hopf bifurcation parameter passes through a critical value. Some computer simulations illustrate correctness of the results.

The multiwavelet research has been an important aspect of the wavelet theory in recent years. This paper summarizes some important properties of multiwavelets. Using the properties of orthogonality and symmetry, A multiwavelet with compact support in [0,1], accurate reconstruction and approximation order 2 is constructed. The multiwavelet has the most advantage of non- boundary distortion. Needless to prefilter, it has better lowpass and highpass characteristics after being balanced. Examples of signal reconstruction and image compression are given, with satisfactory reconstruction results over the single wavelet.

The sidelobe level is an important factor in designing of antenna arrays, and several methods have been developed to improve it. A model of unequally spaced thinned array is given in this paper, which introduces a distance perturbation into the element of a thinned array, in order to improve the sidelobe level. Genetic algorithm and simulated annealing are used to design this kind of array. The pattern of a 200 elements linear array and a 40×20 planar array is presented for the isotropic element pattern, the results verifies the improvement of the sidelobe level.

Image lossless compression plays an important role in the remote sensing field. A new lossless coding algorithm based on perceptron is studied in this paper. The prediction coefficients can be adaptively adjusted by this model. Consequently both spacial and spectral correlation can be reduced. Experiments show that the algorithm is effective for remote sensing image.

A novel variable-rate correlation vector quantizer is presented for image encoding in this paper. Before encoding, four characteristic values of each codeword are computed, then four corresponding sorted codebooks are obtained according to the ascending order of the four characteristic values of codewords respectively. During the encoding, high correlation of the adjacent image blocks and characteristic values of the current processing vector and codewords are considered to encode the current processing vector. Experimental results show that, although the encoding quality of the proposed algorithm is a little worse than that of the conventional VCS, the bit rate of the proposed algorithm is tower than that of the conventional VQ, and the encoding time of the proposed algorithm is much shorter than that of the conventional VCS.

Due to the complexity of automatic target recognition for aerial photo this paper describes an automatic target recognition algorithm based on neural network. It includes two stages, target detection and false-alarm rejection. The first stage operates on the entire image to detect potential targets quickly. The second stage attempts to reject false target-like objects while retaining as many targets as possible. The experimental results show the algorithm is feasible.

The research status of phosphor for FEDs is summarily described. Some key questions are reviewed in details. The CL performance of our newly developed phosphor are reported.

Digital watermarking is a technique, which protects digital product copyright and does not appear until two years ago. And its goal is to protect legal copy and distribution of digital product, it has been the hotspot of international academia. Digital watermarking includes algorithm design and attack technique, which are contrarily promoting each other. In this paper the concept and applications of digital watermarking are briefly discussed, the current various attack methods are fully analyzed and summarized. Several directions of development of digital watermarking at next stage are proposed. These are of great importance to improve and design watermarking algorithm.