This paper describes and explains why image fusion, what is image fusion, and the current research status mainly on wavelet based pixel-based image fusion. Pixel-based image fusion defines the fusion process of original images or the images after pre-processing. Preliminary results of many researches show that the advantages of high-resolution panchromatic image and low-resolution multi-spectral image can be combined by image fusion and the information extraction capability can be improved. The fusion methods evolutes from traditional fusion methods, pyramid based fusion methods to nowadays wavelet based fusion methods. The popular wavelet theory based Mallat algorithm and ¡°¨¢ Trous¡± algorithm are explained. In order to overcome some shortcomings of Mallat algorithm and ¡°¨¢ Trous¡± algorithm, MRAIM algorithm is designed, which is based on the image formation principle and multi-resolution analysis theory. It formulates the Mallat algorithm and ¡°¨¢ Trous¡± algorithm from the theoretical point of view. It can improve the spatial resolution while preserve the hue and saturation unchanged.

A system for target recognition and tracking based on radar and infrared image sensors is presented, which can make use of the complement and redundancy of data from different sensors to improve the precision of target recognition and tracking and the robustness and reliability. For data fusion at characteristic level, characteristics of a target obtained from radar can be used in the IR image-based subsystem to improve the ability of object recognition, and vice versa. The process of target recognition based on IR image analysis is composed of image enhancement, image segmentation and recognition of segmented images. The recognition of segmented objects is divided into two classes: recognition of dot targets and area targets.

The main preoperative way to assess the malignant degree of brain glioma is based on patient's MRI (Magnetic Resonance Imaging) findings and clinical data. Neuroradiologists' experience plays a critical role during the process due to the lack of knowledge about the relationships between influential factors and types of tumor. In order to improve the diagnosis accuracy without incurring the troubles in learning experience from domain experts, we ask help from data mining. Multi-layer perceptron network, decision tree, and rules indication are its common techniques. They are adopted to extract useful diagnostic patterns directly from MRI records of brain glioma patients.

Mining sequential patterns is an important topic in the data mining (DM) or knowledge discovery in database (KDD) research. At present, its research and application are mainly focused on analyzing transaction data of market. As the transaction database is very different from general relational database at the aspect of inner makeup, this paper makes relevant study and discussion on how to mine sequential patterns from general relational database. In this paper, a strategy based on time-window is adopted instead of general Apriori algorithm. At last, an algorithm based on time-window for mining sequential patterns (TW_SP) is put forward, and its correctness and validity is further proved by the experiment in a real relational database.

Color night vision technology is a new kind of night vision means. In this paper, on the base of the study on two-color false color low light level(CLLL) TV technology, the principle and experiment study on single-channel false CLLL TV system are carried out. Deeply, the disadvantages of dual-channel false CLLL TV system are pointed out, LLL image geometric segment technique and spectrum gray-scale compensation technique are researched into. The single-channel CLLL TV system is established. Experiment results show that through the fusion of two spectrum LLL images, the image resolution and the recognizing capability of human eyes can be increased significantly, and the high sensitivity and resolution of single-channel as well as that of the dua-channel technology are realized successfully.

Based on nonstationary random process with variance stationary and mean value with trend, a adaptive weighted fusion estimated algorithm is presented to fuse MMW/IR Combined seeker data in this paper. The nonstationary random process is transformed to the stationary random process by first difference, which is used for estimating measured data variance. Finally, observations are fused through the weighed fusion estimation algorithm. Simulation indicate that this algorithm is simpler, practical and its convergence speed is faster.

The purpose of multispectral image fusion is to merge information from multi-sensor and to improve abilities of information analysis and feature extraction. Discrete wavelet transform can offer a more precise way for image analysis than other multi-resolution analysis. It decomposes an image into low frequency band and high frequency band in different level, and it can also be reconstructed gradually in different level. But this method only decomposes low frequency band in a higher scale, so that it omits some useful details of the images. In this paper, we research an improved discrete wavelet transform. It decomposes high frequency band in higher scale which wavelet analysis does not do. We apply it on image data and give a fusion method in pixel level. Through merging remote sensing image of different wavebands from multi-sensor to a same object by applying method of improved wavelet analysis, we have obtained a fused picture. The method can fuse details of input image successfully, and display information of the each input image perfectly. Comparing with other image fusion methods, satisfactory result has been obtained by applying this method on both objective and subjective performance measure.

In the biological systems genetic information is read, stored, modified, transcribed and translated using the rule of molecular recognition. Every nucleic acid strand carries the capacity to recognize complementary sequences through base paring. Molecular biologists commonly use the DNA probes with known sequence to identify the unknown sequence through hybridization. There are many different detection methods for the hybridization results on a genechip. Fluorescent detection is a conventional method. The data analysis based on the fluorescent images and database establishment is necessary for treatment of such a large-amount obtained from a genechip. The unknown sequence has labeled with fluorescent material. Since the excitation and emission band is not a theoretical narrow band. There is a different in emission windows for different microscope. Therefore the data reading is different for different microscope. We combine two narrow band emission data and take it as two wavelengths from one fluorescence. Which by corresponding UV light excitation after we read the fluorescent intensity distribution of two microscope wavelengths for same hybridization DNA sequence spot, we will use image fusion technology to get best resultsDWe introduce a contrast and aberration correction image fusion method by using discrete wavelet transform to two wavelengths identification microarray biochip. This method includes two parts. First, the multiresolution analysis of the two input images are obtained by the discrete wavelet transform, from the ratio of high frequencies to the low frequency on the corresponding spatial resolution level, the directive contrast can be estimated by selecting the suitable subband signals of each input image. The fused image is reconstructed using the inverse wavelet transform.

An approach to the problem of multimodality medical image fusion is proposed, using 2D discrete wavelet transform (2D DWT). Local image fusion mechanism employing 2D periodic DWT is also presented, which decreases the calculation time and space cost. To evaluate the fusion result, mutual information is employed as a performance measurement. The experiment results demonstrate that the fusion scheme of this study is very well suited for clinical applications, especially for treatment plan when complementary information is needed.

In the area of low light level (LLL) night vision, improving the LLL image quality has received more and more attentions. In this paper, we use the instantaneous laser (near infrared) assistant vision technology to obtain the laser assistant vision (LAV) image and realized the fusion of LLL image and LAV image using wavelet transform. The information feature of the two kinds of images is different because the spectrums they respond are different. In the fusion process, the images are first decomposed based on wavelet transform. We construct the multiresolution analysis of the fused image by considering the multiresolution contrast of the source images. The interested image features can be enhanced. At last the fused image is reconstructed by the inverse wavelet transform. The experiment results show that the fused image is better than any of the individual source images and the fusion of the LAV image and the LLL image can improve the image quality of LLL TV. This technology is meaningful for night vision.

By integrating Fuzzy Kohonen Clustering Network (FKCN) with Fuzzy Dempster-Shafer Evidential Reasoning Theory (FDSERT), a new multi-source data fusion of Remote Sensing information algorithm is proposed in this paper. The new algorithm can be applied in classification of remote sensing image through FKCN learning and FDSERT fusing. Experimental results comparing with the FKCN algorithm indicates that the classification algorithm of multi-source data fusion of Remote Sensing is superior to that of FKCN algorithm. And the algorithm can obviously improve classification accuracy. At the same time, the algorithm can makes the best of expert knowledge. Therefore the algorithm is an effective classification algorithm of Remote Sensing image.

For navigating inside the patient's body, the physician needs to know exactly the 3D position of endoscope probe. We develop a new approach to determine the 3D position of endoscope inside the body using harmless magnetic field, and the pose of endoscope tip by attached a 3D magnetic sensor to endoscope. The experience of this research can be conducted for solving the positioning problem of endoscope inside the patient's body.

This paper describes a method that can be used to track the single moving object automatically in multiple scenes. After preprocessing of the original image sequence, which is sampled from the video camera, the optical flow of the moving object in multiple scenes will be calculated rapidly with the improved algorithm. The moving object region can be extracted completely through several processing of the optical flow field such as region-connection, region- merging, and region-clustering methods. Then, the results of these processing can be used to measure the speed and direction of the moving object quantitatively, and can be used to output the control signal to drive the tracker for tracking the fast-moving object.14

Industrial application of infrared thermography is virtually boundless as it can be used in any situations where there are temperature differences. This technology has particularly been widely used in automotive industry for process evaluation and system design. In this work, thermal image processing technique will be introduced to quantitatively calculate the heat stored in a warm/hot object and consequently, a thermal control system will be proposed to accurately and actively manage the thermal distribution within the object in accordance with the heat calculated from the thermal images.

Some dielectric materials may become birefringent when subjected to an external force, such as an electric field or a mechanical force. For more than a decade, our research group has been engaged in developing 2D optical measurement techniques for the dynamic measurement of charge distributions on a dielectric surface using the electro- optic Pockels effect, the dynamic measurement of electrical field distributions in a liquid using the electro-optic Kerr effect and the measurement of birefringence vector distributions in plastic plates using the photo-elastic effect. The common image processing techniques that are uniquely developed in our work are analyzed and summarized.

Digital navigation concept is introduced to the application of aircraft navigation parameter estimation firstly. In this paper, an estimation method for navigation parameter using digital CCD camera is presented, where navigation parameters represent the position and attitude information of a helicopter for autonomous navigation. The proposed method is composed of relative position estimation and absolute position estimation. Multi-image space orientation is introduced and sequence image (i.e., real-time aerial scene) matching is processed in relative position estimation. Digital image reference map and digital elevation reference map are employed and multi-source matching is processed in absolute position estimation. Integrated position estimation is proposed to perform the estimation of helicopter navigation parameters. Computational simulation and experiment results show that our proposed navigation parameter estimation is feasible for helicopter navigation.

In this paper, a novel multi-scale method is proposed to detect silver grains including more subtle ones in in-site hybridization. The multi-scale representation is built using an undecimated discrete wavelet transform, a biorthogonal B- spline wavelet basis is applied to the transform. A multi- scale and orthogonal feature set can be acquired from the wavelet decomposition as input to a multilayer feed-forward neural network which maximizes the separation between the presence and absence of grains. The resulting map of the classification indicates the presence and location of silver grains. We use it to restrict enhancement to highly localized regions identified by the detection algorithm. Then an inverse wavelet transform is applied to reconstruct the detected and enhanced objects. Experiment results show that the proposed approach is able to highlight silver grains while significantly reducing the contrast of the remaining image.

For vision-based controlled flexible arm, the camera and driving motor are located at different locations, that is, noncollocated control system. The stability will not be guaranteed. Therefore, simplified finite dimensional dynamic models of flexible arm will be developed for the purpose of controller design based on Kane¡¯s equations and Euler-Berneuli theory. This paper addresses the issue of selecting the shape and mode in developing finite order models and designs suitable PID controller for a flexible arm with vision-based control. Results show that designed controller can match the requirements for vision-based controlled flexible arm when the influence of the dead band will be eliminated.

There is an extensive application requirement of airborne 3D imager in city planning today. An airborne 3D imager has been developed and it is introduced first in this paper. Some solutions in the imager to improve the efficiency and accuracy of 3D imaging for city planning are specially analyzed. After some flight missions of the imager for city planning it provided lots of data of ¡®digital city¡¯. Through information and image processing a series of images with a scale of 1:10000 are produced. Some typical images and 3D city models of several Chinese cities are given finally.

System CASA (computer-assisted semen analysis) is a medical applicable system which gets the sperm motility and its parameters using image processing method. But there is no any authoritative administration or academic organization gives a set of criterion for CASA now result in lowering the effective compare of work between the labs or researchers. The average path and parameters relative to it as average path velocity, amplitude of lateral head displacement and beat cross frequency are often unable to compare between systems because of different algorithm. The paper presents a new algorithm that could define the average path uniquely and compute those 3 parameters above quickly and handy from any real path.

The traditional methods to implement image processing are either using general programmable digital signal processor (DSP) based instruction or using dedicated signal processing chips. The dedicated chips can deal with tasks they are designed for efficiently, but with others tasks their performance dropped very mich. The general processors have high flexibility but low performance. In many designs, we usually take multi-DSP system to attain the high performance and flexibility but we must face the complex software task on multi-processor system. In this paper, we introduce the concept and advantage of reconfigurable technology. Then we analyzed the real-time requirements of some image-processing algorithms.

To predict image features in an image from image features in two other images has a wide potential appliance in computer vision such as Visual Recognition, and model-based vision animation and view synthesis, object detection and tracking. In this paper, we provide a novel method to predict image features in the third image based on the trifocal tensor.

The Very Long Baseline Interferometry (VLBI) technique uses small radio telescopes to construct a huge synthetic telescope through earth rotation, resolving details as fine as 0.1 milliarcsecond. This paper introduces the principle of the VLBI synthesis and presents the mathematic model of the powerful high-speed FX type Chinese VLBI Network correlator. The principle of the delay and phase compensation, the signal transformation in the correlator, and the final complex visibility function are also presented, as well as the correlator's block diagram. The model is great helpful in designing and adjusting the FX VLBI correlator. Examples demonstrate how to use the model to judge the output visibility function and analyze the observing station performance and the astronomical line spectrum.

Prevention and early diagnosis of tumors in mammogram are foremost. Unfortunately, these images are often corrupted by the noise due to the film noise and the background texture of the images, which did not allow isolation of the target information from the background noise, and often results in the suspicious area to be analyzed inaccurately. In order to achieve more accurate detection and segmentation tumors, the quality of the images need to improve, (including to suppressing noise and enhancing the contrast of the image). This paper presents a new adaptive histogram threshold method approach for segmentation of suspicious mass regions in digitized images. The method use multi-scale wavelet decomposition and a threshold selection criterion based on a transformed image¡¯s histogram. This separation can help eliminate background noise and discriminates against objects of different size and shape. The tumors are extracted by used an adaptively bayesian classifier. We demonstrate that the method proposed can greatly improve the accuracy of detection in tumors.

This paper presents an alternative algorithm for number plate recognition. The algorithm consists of three modules. Respectively, they are number plate location module, character segmentation module and character recognition module. Number plate location module extracts the number plate from the detected car image by analyzing the color and the texture properties. Different from most license plate location methods, the algorithm has fewer limits to the car size, the car position in the image and the image background. Character segmentation module applies connected region algorithm both to eliminate noise points and to segment characters. Touching characters and broken characters can be processed correctly. Character recognition module recognizes characters with HHIC (Hierarchical Hybrid Integrated Classifier). The system has been tested with 100 images obtained from crossroad and parking lot, etc, where the cars have different size, position, background and illumination. Successful recognition rate is about 92%. The average processing time is 1.2 second.

Position and Orientation System/Direct Georeferencing (POS/DG) data is important to hyperspectral images, for it has much information of flight attitude, such as absolute position (x,y,z) and rotation parameters. The largest advantage of this method, using POS/DG data to correct remote sensing images, is to save much manual work and money. This method doesn't need any ground work. It can get high quality images just through a correction program. It is more economical, simpler and faster than conventional methods. This article concentrates on the research of how to use POS/DG data to correct the hyperspectral images (in this article is PHI images) and discuss the efficiencies of a few kinds of resampling methods.

The objective of AVLPRS is to accurately extract and recognize the vehicle number-plate from the complicated background. A practical AVLPRS is developed which can realize functions of real-time image capturing, pre-processing, number plate segmentation and character recognition.

In this paper, the filamentary pattern of dielectric barrier discharge has been processed by using Photoshop, the coordinates of each filament can also be obtained. By using Photoshop two different ways have been used to analyze the spatial order of the pattern formation in dielectric barrier discharge. The results show that the distance of the neighbor filaments at U equals 14 kV and d equals 0.9 mm is about 1.8 mm. In the scope of the experimental error, the results from the two different methods are similar.

This article qualitatively puts forward the uncertainty principle of electro-optical system vulnerability to ECM, a new objective function of EOSVECM, namely the turning point of dominance of utilizing electromagnetic spectrum is pointed out and analyzed, independent and divergent straight-line reasoning rule of dynamics principle used for evaluating EOSVECM quantitatively now, such as maximum- likelihood reasoning, is expanded into systemic and convergent mixed reasoning rule, such as aggregated transduction common-sense reasoning.

Under the direction of the principle of EOSVECM, a new semi- quantitative index--the turning point of the dominance of utilizing electromagnetic spectrum--is put forward in this article as the first objective function of EOSVECM, and 18 attributes (connotation) of this kind of objective function is analyzed, so as to inaugurate the concrete approach to this kind index of EOSVECM.

This paper describes a QoS guarantee architecture suited for best-effort environments, based on ideas from microeconomics and non-cooperative game theory. First, an analytic model is developed for the study of the resource allocation in the Internet. Then we show that with a simple pricing mechanism (from network implementation and users' points-of-view), we were able to provide QoS guarantee at per flow level without resource allocation or complicated scheduling mechanisms or maintaining per flow state in the core network. Unlike the previous work on this area, we extend the basic model to support inelastic applications which require minimum bandwidth guarantees for a given time period by introducing derivative market.

The light spot images on the coating rough surface with the plasma-thermal sprayed coating materials Z_rO₂8Y₂O₃ and N_iC_rAlY when the laser beam (visible, wavelength is 632.8 nm, power is 2 mw) with different diameter shined on it have been studied and its experimental demonstration has also been presented in this paper. The Charge Coupled Device camera is applied to detect the light spot image field distribution, and 3D distribution of the light spot images have been obtained by computer. Above experiment results give us some basis knowledge to measure the coating thickness in the next step, which shows the physical behavior of the light spot. In addition, the different profiles of Gauss laser spot image on the rough surface for Z_rO₂8Y₂O₃ and N_iC_rAlY coating have been analyzed.

With the development of remote sensing data acquisition technique, more and more image data have been acquired. To process the explodely increasing data, data fusion technique has been used widely. Compared with traditional methods such as IHS, PCA, HPF, etc, multiresolution analysis data fusion methods can preserve spectral character much better. There've been many multiresolution-based data fusion methods so far, moreover, wavelet-based and pyramid-based methods have played important roles in data fusion area. However, traditional wavelet-based fusion methods always decompose every multisource or multitemporal image into several parts as low frequency and high frequency, and then continue the decomposition to the low frequency part. Afterwards, fuse the same levels low and high frequency parts of different images, and then reconstruct the fused image. Now a novel wavelet transform-based fusion method is described, which uses a wavelet package to decompose the multisource or multitemporal images at either low or high frequency parts. Then, at the same level, utilizes a threshold and weight algorithm to fuse the corresponding low frequency parts, at the same time applies Lis high-pass filter on fusing the high frequency parts. At last, experiment is performed on two multitemporal images to validate this method.