An extension to the like-value procedure is proposed for calibrating multitemporal sequences of TM and ETM+ imagery to absolute reflectance values. The procedure described differs from the conventional approach by using a calibrated reference image and constraining the line of best fit to pass through the modeled digital number (DN) for a target with zero surface flectance. It was demonstrated that the resulting lines calibrate image DNs to surface reflectance values by simultaneously correcting for the additive and multiplicative components of the sensor output that result from changes in sensor response, solar position and atmospheric effects over time. In a comparison to reference image reflectances, the precision of the procedure was shown to be very good for the visible spectral bands but poorer for the infrared bands. Generally errors in retrieval were within the range considered acceptable for most practical applications. For the data used in the analysis, the procedure was shown to be insensitive to uncertainty in input parameters required to model path DN.

This article introduces the processing procedure and key techniques of the CBERS-1 IRMSS 0-level image processing system, includes pre-processing, radiation correction, and geometry correction. The main purpose of the processing is that using image engineering method to remove the influence comes from camera's physical feature so as to improve the image quality.

Remote-sensing image fusion is becoming widely used in object recognition because of complementary nature of images from different sensors. In the past few years, many fusion methods have been introduced, where intensity-hue-saturation transform (IHST)-based fusion method, wavelet transform (WT)-based fusion method and high pass filter (HPF)-based fusion method are commonly used. There are complementarities between the three fusion methods, so this paper presents a new fusion method based on WT, IHST and HPF, which overcomes the shortcomings of the three fusion methods and exploits their strong points. Our new method uses WT-based fusion method to preserve the spectral information of the original multispectral image, and IHST-based fusion method to improve spatial presentation of the fused image, and HPF-based method to merge the low-frequency part of the panchromatic image with that of the I channel data of the multispectral image in WT domain to enhance spatial information and avoid block effect in WT-based fusion method. Experiment results show that our new method is efficient. It not only preserves the spectral information of the original multispectral image very well, but also enhances spatial presentation of the fused image largely.

We attempted a method of fusing a coarse resolution hyperspectral image and a high spatial resolution image with a few spectral bands to produce a high resolution hyperspectral image, and to extract the spectra of the end-members. The method is based on the linear spectral mixing model and an iterative maximum-likelihood algorithm is used to invert the mixing equation. The effects of noise and misregistration error are investigated. Misregistration seems to be a main factor determining the accuracy of the final products.

Advances in sensor technology for Earth observation make it possible to collect multispectral data in much higher dimensionality. Such high dimensional data will it possible to classify more classes. However, it will also have several impacts on processing technology. First, because of its huge data, more processing power will be needed to process such high dimensional data. Second, because of its high dimensionality and the limited training samples, it is very difficult for Bayes method to estimate the parameters accurately. So the classification accuracy cannot be high enough. Neural Network is an intelligent signal processing method. MLFNN (Multi-Layer Feedforward Neural Network) directly learn from training samples and the probability model needs not to be estimated, the classification may be conducted through neural network fusion of multispectral images. The latent information about different classes can be extracted from training samples by MLFNN. However, because of the huge data and high dimensionality, MLFNN will face some serious difficulties: (1) There are many local minimal points in the error surface of MLFNN; (2) Over-fitting phenomena. These two difficulties depress the classification accuracy and generalization performance of MLFNN. In order to overcome these difficulties, the author proposed DPFNN (Double Parallel Feedforward Neural Networks) used to classify the high dimensional multispectral images. The model and learning algorithm of DPFNN with strong generalization performance are proposed, with emphases on the regularization of output weights and improvement of the generalization performance of DPFNN. As DPFNN is composed of MLFNN and SLFNN (Single-Layer Feedforward Neural Network), it has the advantages of MLFNN and SLFNN: (1) Good nonlinear mapping capability; (2) High learning speed for linear-like problem. Experimental results with generated data, 64-band practical multispectral images and 220-band multispectral images show that the new algorithm can overcome the over-fitting phenomena effectively and improve the generalization performance of DPFNN greatly. The classification accuracy of DPFNN with the new learning algorithm is much better than the traditional one.

Conventional classification methods cannot recognize the phenomena of "same spectrum with different land matters" so as to degrade classification accuracy. To solve the problem, this paper proposes a new classification method using neural network based on generalized image, where the space information of the image are exploited. Firstly, we combine the original image with its smoothed image to form a binary set called as a "generalized image," which contains the space information of the original image. Secondly, we make use of artificial neural networks (ANN) to train and classify the "generalized image." Finally, we get the classification result of the original image from that of the "generalized image." Experiment results show that the new method is very efficient, and the classification accuracy is improved largely compared with the classic ANN method.

Remote sensing images acquired by the sensors at platforms near land surface, airplane and satellite, usually have large volume and miscellaneous data formats. So it is not feasible for the users to browse remote sensing images and evaluate the quality of images and select the suitable images on Internet. Moreover, it is inefficient to read and transfer remote sensing images real-timely in a standard image viewer due to their miscellaneous data formats. In order to clear up the problems, the metadata and microimage are extracted from various remote sensing images, managed by the database management system software, and browsed and evaluated on Internet to decide which images are the real wanted. The process of working includes the 4 steps (1) Create metadata for the remote sensing images. The metadata consist of image data format, longitude and latitude of image range, data and time, spatial resolution, sensor attributes (field of view, bands, performance and precision etc), platform attributes (stand near land surface, airplane or satellite), flight path or orbit attributes of aerial and space observation etc. (2) Create microimage for remote sensing image. Firstly, the remote sensing images are projected to the same coordinate system by the geometric correction, so all images can be matched correctly. Then the microimages are built through 1:10 or 1:5 cubic convolution sampling the corrected images. (3) Build a database to store and manage the metadata and microimages, and create pointers to hyperlink the remote sensing images self. (4) Develop the browse interface, publish the remote sensing image base on Internet, and receive the users' order forms. The wanted images will be sent on CDROM if the orders are accepted. The interface is visualized. Here, a color spectrum is used to express the bands. A clock is for time and landscape is for days in one year. And place is located by moving your mouse on the map. The pixel sizes are shown through levels on a pyramid. By this metadata and microimage approach, the remote sensing images can be browsed, evaluated and ordered on Internet conveniently. It is feasible way to manage the remote sensing images.

In this paper, the authors proposed a new theory and method, the phase-separation analysis of remote sensing information field of metallogenetic environment for non-model ore deposit prediction. The theory of nonmodel ore deposit prediction suggests that the forming of ore deposit result from multiple changes of many geological factors; and in a particularly geological environment, the places where tectonic movement of multi-times piled up together probably produce large or giant type deposits; and the existence of great mineralized body maybe lead to some remarkable differences in composition, in structure, in geophysical field and in geochemical field from the surrounding geological background, even lead to anomalies of biosphere and atmosphere of the earth. Therefore, on the basis of geology and other data, combining RS with GIS, and through decomposing multivariate information fields, feature extracting and seeking anomaly, it is feasible to establish the natural models of ore source bodies to predict related ore deposits correctly.

A weighted neighbour intensity interpolation for resampling of remote sensing imagery has been developed. Examples of the resampling of SeaStar SeaWiFS images by the interpolation are presented in this paper. The weighted neighbour intensity interpolation has been compared with other intensity interpolations. Advantages and disadvantages of the weighted neighbour intensity interpolation over others have been discussed.

In this paper, we propose a novel and automated line extraction algorithm in multi-spectral images, which fully utilize the complementary information among multi-spectral images. It consists of three main aspects: Firstly, edges are extracted from every spectral image. Then, the edge points from all spectral images are grouped into combined line-support regions according to certain fusion rules. Finally, fits the regions and generates the fused lines. The new algorithm is applied to some real multi-spectral images. The experimental results show that the new algorithm is effective.

The benefits of EO-1 data, and especially Hyperion hyperspectral data, are being studied at sites in the Coleambally Irrigation Area of Australia where a seasonal time series has been developed. Hyperion can provide effective measures of agricultural performance through the use of spectral indices if systematic and random noise is managed and such noise management methods have been established for Coleambally. Among the sources of noise specific to Hyperion is the spectral “smile” which affects the location of the red-edge -- an important index in agricultural assessment. We show how this phenomenon, which arises from the pushbroom technology of Hyperion, affects the data and discuss how its effects can be overcome to provide stable and accurate measures of the red-edge and related indices. HyMap airborne data are used to evaluate the results of the methods studied. This paper also shows how future pushbroom instruments should consider the wavelength sampling step in their design if it is intended to remove the “smile” effects by a systematic software processing.

This paper introduces a novel methodology for texture object detection using genetic algorithms. The method employs a kind of high performance detection filter defined as 2D masks, which are derived using genetic algorithm operating. The population of filters iteratively evaluated according to a statistical performance index corresponding to object detection ability, and evolves into an optimal filter using the evolution principles of genetic search. Experimental results of texture object detection in high resolution satellite images are presented to illustrate the merit and feasibility of the proposed method.

The seasonally flooded forest and upland sites in the Kakadu National Park, near Jabiru, Northern Territories, Australia were the site of extensive field measurements, bird community observations and airborne remote sensing during an initial NASA/Jet Propulsion Lab AIRSAR campaign in 1993, a field visit in 1994 and combined remote sensing and field activities during the PACRIM I Project in fall 1996. The overarching purpose of these studies was to use remote sensing technology as a way to extend intensive avian biodiversity and census field observations, as well as structural vegetation measurements from a limited survey area to the regional scale. During these two visits to the Kakadu area, field measurements were made within the dominant forest types in this region, primarily mixed Eucalyptus sp. woodlands, and open- and closed-forest sites dominated by Melaleuca sp. across a range of dry to perennially-flooded sites. Bird community measurements showed vegetation structure is needed to understand habitat relationships. A major vegetation difference between the two years was an increase of 2-3 times in leaf area index at comparison sites from 1994 to 1996. The greatest LAI at any site was 2.52 in the wet Melaleuca site near Munmalary in 1994.

High-resolution satellite imagery has become widely available recently and it enables urban remote sensing to not only classify land-use, but also map the details in urban environment. However, due to high object density and scene complexity, normally it is extremely difficult to automatically extract urban objects solely based on images. This paper describes our approach to detect buildings by fusing high-resolution IKONOS satellite images and airborne laser scanning data. With the high spatial resolution, rich spectral signature of IKONOS images and the very accurate positioning information of laser data, our data fusion methods show an efficient way to exploit the complementary characteristics of these two kinds of dataset for the purpose of building detection. In order to simplify the complexity of processing, a top to down strategy is generally applied to extract features of objects from coarsely to finely, and multiple cues are also derived and fused at different processing levels. The paper describes the developed framework and experimental results in detail, and also discusses both the advantage and deficiencies of the approach.

SAR image segmentation can be converted to a clustering problem in which pixels or small patches are grouped together based on local feature information. In this paper, we present a novel framework for segmentation. The segmentation goal is achieved by unsupervised clustering upon characteristic descriptors extracted from local patches. The mixture model of characteristic descriptor, which combines intensity and texture feature, is investigated. The unsupervised algorithm is derived from the recently proposed Skeleton-Based Data Labeling method. Skeletons are constructed as prototypes of clusters to represent arbitrary latent structures in image data. Segmentation using Skeleton-Based Fuzzy Clustering is able to detect the types of surfaces appeared in SAR images automatically without any user input.

In this paper, we propose a pixel-based image fusion algorithm that combines the gray-level image fusion method with the false color mapping. This algorithm integrates two gray-level images presenting different sensor modalities or at different frequencies and produces a fused false-color image. The resulting image has higher information content than each of the original images. The objects in the fused color image are easy to be recognized. This algorithm has three steps: first, obtaining the fused gray-level image of two original images; second, giving the generalized high-boost filtering images between fused gray-level image and two source images respectively; third, generating the fused false-color image. We use the hybrid averaging and selection fusion method to obtain the fused gray-level image. The fused gray-level image will provide better details than two original images and reduce noise at the same time. But the fused gray-level image can't contain all detail information in two source images. At the same time, the details in gray-level image cannot be discerned as easy as in a color image. So a color fused image is necessary. In order to create color variation and enhance details in the final fusion image, we produce three generalized high-boost filtering images. These three images are displayed through red, green and blue channel respectively. A fused color image is produced finally. This method is used to fuse two SAR images acquired on the San Francisco area (California, USA). The result shows that fused false-color image enhances the visibility of certain details. The resolution of the final false-color image is the same as the resolution of the input images.

This paper deals with the registration of images which are subject to translation, rotation, scaling and other geometric transformations. The problem is tackled with a contour-based registration technique. We present a novel approach for finding a representation of planar curves and matching such two representations in this paper. In our framework, the matching of curves is proceeded by (1) using connected equi-length line segments (CELLS) to represent curves, (2) attaching an identification vector to each line segment, which reflects the distribution of the rest of line segments with respect to the current one using orientation difference between cells. A new matrix called Orientation Difference Matrix (ODM) has been constructed from the identification vectors. This approach uniquely specifies a curve and the representation for matching is invariant under rotation, scaling and translation of the curve. A practical use of the proposed approach is demonstrated by registering a SAR image of a certain area to a map.

Data fusion of SAR (during flood event) and LANDSAT TM (before flood event) has been well known to detect flood disaster, owing to that the all-weather and all-time ability of microwave remote sensing and the abundant information and good interpretation of TM. In this paper, firstly, an approach named TIN (Triangulated Irregular Network) was used for registration, with better accuracy than other general methods. Secondly, according to statistic analysis including interband correlation and entropy, etc. the selection of the optimal bands of TM was discussed. In order to monitor flood event better and more quickly in future, we analyzed various fusion approaches to see which is optimal and less time-consuming with the data of 98's flood happened to the Yangtze River. As a result, the pseudocolor image of composition of TM5-SAR-TM3 is the best, however, the combinations of TM5-SAR-TM2, TM7-SAR-TM2 and TM7-SAR-TM3 are also alternative and feasible.

The multiplicative nature of the speckle noise in SAR images has been a big problem in SAR image segmentation. The maximum likelihood region merging segmentation algorithm is researched and employed in this paper. And an adaptive edge detect procedure is introduced into the algorithm here to obtain edge image before the region merging processing, so that the edge iamge can be the guidance of the primary segmentation. Before theoretic analysis and segmentation experiment show the better segmentation performance of the improved algorithm with the adaptive edge detect involved in.

In order to research the global change and analyze the characteristics of land surface or atmosphere, an Earth Observation Station usually needs a high-definition large screen to display the satellite remote sensing image. Under the development of digital imaging technology, this application can be put into practice. In this paper, the optical engine, a key technique in the digital imaging, is introduced. Then the basic principle and technical difficulties of optical engine are discussed in detail. It is testified that the optical engine technique can make the satellite remote sensing image displayed in high-definition model.

The development of the remotely sensed techniques enlarges the applications of the remote sensing imagery. The clustering of high resolution imagery is difficult, due to the fact that the minor objects, such as roads, make the appearance of the same category region non-uniform. This paper proposes a new approach to cluster high resolution remote sensing imagery. The new clustering approach includes three steps as the following: Firstly, eliminate the minor components in the moving windows. Secondly, compute the image features, such as the energy, some high order cumulants and central moments of pixels' values in moving windows. Lastly, apply the BPC neural network, which is combined by a Back-Propagation (BP) neural network and a Competive neural network, to cluster images according to the image features. Two methods, minimum distance method and the K-means method, are compared with the new clustering approach, proposed by this paper, by using SPOT images for clustering residential areas and agricultural areas in the suburbs of Beijing. The experimental results show that the new clustering approach has the higher clustering accuracy.

Good detection, good localization and single response to a single edge are the three performance criteria for edge detection. Based on the mathematical formulations about detection, localization and single response criterion, the paper explains the theory of how to attain the optimal edge detection operator in details. The procedure of designing edge detection operators for any shape edges is introduced by illustrating how to attain the one dimensional step edge optimal detection operator. Based on the three performance criteria, a new edge detector is proposed, which is named as sine-operator. The sine-operator outperforms the first derivative of Gauss function according to the above three criteria. The results show that the sine-operator proposed by the paper can obtain good performance for both one dimensional signal and two dimensional image with noise.

In this paper, a novel hierarchical image fusion scheme based on wavelet multi-scale decomposition is presented. The basic idea is to perform a wavelet multi-scale decomposition of each source image first, then the wavelet coefficients of the fused image is constructed using region-based selection and weighted operators according to different fusion rules, finally the fused image is obtained by taking inverse wavelet transform. This approach has been successfully used in image fusion. In addition, with the use of the parameters such as entropy, cross entropy, mutual information, root mean square error, peak-to-peak signal-to-noise ratio, the performance of the fusion scheme is evaluated and analyzed. The experimental results show that the fusion scheme is effectual.

D-S evidence theory is a useful method in dealing with uncertainty problems, but its application is limited because of the shortcomings of its combination rule. The number of its multiplication increases by exponential principle as evidences grow. In addition, the combination results may be unacceptable if the evidences involved conflict one another highly. So the paper introduces min-max operator with comparison operation instead of multiplication operation and addition operation in the combination rule. In order to combine highly conflicting evidences effectively, the evidences' conflicting probability is distributed to every proposition according to its average supported degree. The new combination rule saves the computing cost largely and improves the reliability and rationality of combination results. Although evidences conflict one another highly, good combination results are also obtained.

The central idea of this paper is that image data mining could be performed directly on the 2D image representation, by applying some scan techniques on the 2D image, which are different than the raster scan. In this paper, we present the comparison of spatial data sets using bit sequential format on a unique vector form which converts between one quadrant tree and some sub-quadrant trees. Then, we describe how the bit-vector might be used to generate the associations among scan patterns in which when some object attributes are extracted in a data process, the others are extracted too.

This paper presents a line extraction algorithm in SAR (Synthetic Aperture Radar) images. The algorithm is designed based on the statistical characteristics of the speckle in SAR image. Three steps are involved. Firstly, a new edge detector, which combines the Canny operator and Ratio operator, is used to detect the edge points and calculate their directions, then the edge points are grouped according to their edge direction to form the initial lines. Finally, a high-level grouping step connects the fragmental lines. The proposed new edge operator is CFAR (Constant False Alert Rate) and prevents the line from cleavage. The algorithm has been applied in the X-band airborne SAR images, and the results are presented at the end of this paper.

Automatic extracting and updating road networks is a key work for updating geo-spatial information especially in developing countries. In this paper, a whole framework for automatic road extraction is presented firstly. Then the strategy and algorithms using GIS data for road extraction are discussed. A hybrid method based on structure information and statistical information for road extraction is emphasized in this paper. Different extraction strategy and grouping techniques are employed for different extracting methods. Because of the importance of structure information in road extraction, the extraction of candidate road segments based on structure information is described. For road extraction from images with different resolution based on structure information, different grouping technique is applied. The grouping technique based on whole relation and the grouping technique based on new profile tracing algorithm is separately employed for images with low resolution and with high resolution. The road extraction based on statistical information is the supplement of structure information. A new statistical model is presented and the candidate road-tracing algorithm based on adaptive template is discussed. And the grouping based on ribbon-snake model is briefly introduced. Automatic road recognition is a necessary task for automatic extracting road networks. So aiming at this we put all kinds of road recognition knowledge into the knowledge base and build a road recognition expert system. The fuzzy theory is applied for representing road models and road knowledge reasoning. The strategy for using global information to guide the further road extraction is presented. At last some examples and the summary are given.

This paper describes a Web-based server software that provides subsetting, resampling, georectificationon, and reformatting for standard data products of the Earth Observing System (EOS) program initiated by U.S. National Aeronautics and Space Administration (NASA). Designed upon standardized interface specifications, the server allows clients to access EOS data in interoperable, personalized, on-demand manners, facilitating the use of NASA EOS data in research and application communities.

As we know that morphology is the technique that based upon set theory and it can be used for binary image processing and gray image processing. The principle and the geometrical meaning of morphological boundary detecting for image were discussed in this paper, and the selecting of structure element was analyzed. Comparison was made between morphological boundary detecting and traditional boundary detecting method, conclusion that morphological boundary detecting method has better compatibility and anti-interference capability was reached. The method was also used for L.V. cineangiograms processing. In this paper we hoped to build up a foundation for automatic detection of L.V. contours based on the features of L.V. cineangiograms and Morphological theory, for the further study of L.V. wall motion abnormalities, because wall motion abnormalities of L.V. due to myocardia ischeamia caused by coronary atherosclerosis is a significant feature of Atherosclerotic coronary heart disease (CHD). An algorithm that based on morphology for L.V. contours extracting was developed in this paper.

In many case, the traditional image geometric calibration method, i.e. polynomial warping will not accommodate the precision transformation required. So the least square method is often used to improve the precision. But because of the multi-relation, the calibration is not so good as expected. In this paper, the theory and experiment studies on the traditional least square method used in image calibration are carried out systematically. With regard to the diversity of image distortion, the selected variables method is applied, which smartly analyses the variables of transformation equation of different images. Through reducing the variables unnecessary, the multi-relation is reduced. Then the ridge-regression is employed, in which the biased estimation is used to solve the huge confidence interval problem that is resulted by multi-relation. The theory and experiment results show that by using this new technique, the multi-relation is reduced and the precision is improved apparently.

According to characteristic of image wavelet transform and interpolation, this paper proposes a remote sensing image interpolation method combining wavelet transform and interpolation algorithm, which can improve the remote sensing image resolution. Experiments show that the algorithm can properly retain abundant high frequency information in original remote sensing image. After interpolation processing and wavelet reconstruction, we can obtain a remote sensing image with higher resolution, better visual effect, higher Signal Noise Ratio (SNR), more detail information and no apparent warp. Therefore, this algorithm is an effective method of super-resolution remote sensing image processing.

Based on lifting scheme and the construction theorem of the integer Haar wavelet and biorthogonal wavelet, we propose a new integer wavelet transform construct method on the basis of lift scheme after introduciton of constructing specific-demand biorthogonal wavelet transform using Harr wavelet and Lazy wavelet. In this paper, we represent the method and algorithm of the lifting scheme, and we also give mathematical formulation on this method and experimental results as well.

The purpose of the project is to find information of LUCC (land use and land cover change) from 1985 to 2000 interval two periods in study area. Such as Changjiang delta, Shiyanghe drainage, Huanghuaihai district, these results will be provided to build LUCC database in our country next project. The remote sensing main data include ETM+, TM and parts of SPOT, because study area is very extensive, time is short too, it is very important to explore and develop extraction methods of LUCC information. Through making some experiments, the author finds some methods effectively base on study area as follows: (1) Methods of change detection efficiently are change identification methods in spectral feature, such as Image Subtraction and Threshold, PCA, Multi-temporal false color composite, bands ratio and so on. (2) Using ancillary data. Remote sensing allows large area to be monitored quickly, but GIS information provide source of very valuable information, which has great advantages over direct field surveying. Such as DEM, land use special subject information, vegetation cover. The use of these information, and remote sensing can together greatly attain changed information. To build new decision tree improve the efficiently of land use/land cover analyses too.

In this paper a new pattern or feature abstraction algorithm was developed simulating our human eyes scan on situation leapingly and focus its attention only to limited space, in order to make a physiognomic analysis of the remote sensing images of the earth's surface, in the end to acquire a description like that in where the foothill, the forest, the desert or the alluvial pie slice was, and so on. At first a size-changeable and edge-fuzzy window was designed to get many samples of the earth's surface through sliding around the image, all these samples served for the learning of a Support Vector Machine model, which was designed to make pattern's classifications. This process was repeated in different area, with different sampling size, to different pattern and lasting different times. Once some distinct local patterns were found and mastered, a self-organizing of comparability assembling will happen based on the similarity of some types of local patterns to form a holistic description or understanding of the remote sensing image. Our aim was to compartmentalize the image by physiognomic features. At the end of this paper the results of classification experiment and application of this method to some actual visible light images were presented. This method was suitable to extend to other pattern recognition problems with texture property.

This paper provides a useful method to derive coastline information from TM/ETM imagery. When the accurate cartography data of study district are rare or out of date, the data of remote sensing (RS) are further more important. It gives the data processing workflow and the necessary working steps to be taken in the processing of coastline fetching. The coastline information in Fujian province of China is derived using this method and it fits the 1:100000 scale map very well. At last, as an argument, a kind of high resolution RS data, IKNOS imagery is also presented, which will be mainly used in the coastline information fetching in the future.

In order to effectively reduce the speckle noise of SAR image without blurring the useful information, this paper carry out a new way to use better WaveShrink function and better way to use edge information. Our proposed way mainly includes two steps: (1) Getting the SAR images edge information by using wavelet; (2) Using the 1-Dimension Garrote WaveShrink technique based on the edge information to reduce the speckle noise. By using the one-order derivative of Gaussian function and setting a proper dilation scale factor we can detect the edge without the noise. By introducing the Hysteresis threshold technique used by Canny we can let the edge has better connectivity. We using a group of Donoho signal with additive noise to prove the Garrote WaveShrink has better performance than the soft and hard function. In this paper, we use the 1-Dimension WaveShrink technique, so we can more easy to choose the threshold than use 2-Dimension technique. At the end of this paper the comparisons of the proposed way with the 2-Dimension WaveShrink way and the classical Lee filter show very good results.

Change detection is a key topic in land use/land cover related studies and significant efforts have been made in the development of methods for change detection. In this article a multivariate analysis method based on canonical transformation is introduced into change detection using multi-temporal remote sensing imageries. Afterwards an automatic unsupervised discriminating technique based on the Bayes Rule of Minimum Error is employed for changed areas identification in the difference image. Experimental results of a case study using Landsat TM imageries are presented to demonstrate the effectiveness of our method.

In this paper, a new classified method for satellite sensing data is proposed by using both a neural network and knowledge reasoning technique. Neural network model can better distinguish land of level I, but if more subdivision needed, it seldom get satisfied result. Knowledge reasoning system can use human geographical knowledge to improve the classification results, but it needs a large amount of assistant knowledge to classify the data correctly. The new method makes use of the advantages of both the neural network and knowledge reasoning technique, and fulfils layered intelligent extraction of linear object and plane-like object for satellite sensing image. It firstly extracts water and road information by neural network and pixel-based knowledge post-processing method, then remove them from original image, and then segments other plane-like object by neural network model, and respectively computes their features, including texture, elevation, slope, shape etc., then extracts them by polygon-based uncertain reasoning method. At last experimental results indicates that the new method outperforms the single neural network method and moreover avoids the complexity of single knowledge reasoning technique.

While using conventional two-dimensional wavelet transform for texture analysis and classification the image decomposition is carried out with separable filtering along the abscissa and ordinate using the same pyramidal algorithm as in the one-dimensional case. This process is simple and can be implemented easily in practical applications, however, it is rotation-sensitive and some information may be lost since the decomposition is performed only in low frequency channels. In this paper the quincunx transform using nonseparable sampling and filters is substituted for conventional dyadic transform. Since the energy of natural textures is mainly concentrated in the mid-frequencies, this transform can preserve more of the original signal energy and can provide more reliable description of the texture. At the same time, the tree-structured wavelet transform or wavelet packets is applied instead of using the pyramid-structured one. With this transform, we are able to zoom into any desired frequency channels for further decomposition and a series of subimages with the largest energy can be obtained for an image. In comparison with conventional wavelet transform, it can be concluded that this transform can still reach higher classification accuracy especially for the characterization of noisy data.

A new hyperspectral image compression algorithm-NMST (Near Min Spanning Tree) is proposed. The near minimum spanning tree is constructed according to the image structure and is taken as a prediction tree in image compression. The result shows the NMST algorithm can improve the compression speed with little decrease of compression ratio.