. We present a method for change detection in images using Conditional Adversarial Network approach. The original network architecture based on pix2pix is proposed and evaluated for difference map creation. The paper address three types of experiments: change detection in synthetic images without objects relative shift, change detection in synthetic images with small relative shift of objects, and change detection in real season-varying remote sensing images.

/pdf/change-detection-in-remote-sensing-images-using-conditional-2n53ntps1b.pdf

Change detection in remote sensing images using conditional adversarial networks

In this paper, we propose a probabilistic model for detecting relevant changes in registered aerial image pairs taken with the time differences of several years and in different seasonal conditions. The introduced approach, called the conditional mixed Markov model, is a combination of a mixed Markov model and a conditionally independent random field of signals. The model integrates global intensity statistics with local correlation and contrast features. A global energy optimization process ensures simultaneously optimal local feature selection and smooth observation-consistent segmentation. Validation is given on real aerial image sets provided by the Hungarian Institute of Geodesy, Cartography and Remote Sensing and Google Earth.

/pdf/change-detection-in-optical-aerial-images-by-a-multilayer-1hjdeslo8w.pdf

Change Detection in Optical Aerial Images by a Multilayer Conditional Mixed Markov Model

A system that includes a desk top assembly of a display and sensors mounted on a robotic arm. The arm moves the assembly so that it remains within position and orientation tolerances relative to the user's head as the user looks around. Near-field speaker arrays supply audio and a microphone array senses a user's voice. Filters are applied to head motion to reduce latency for arm's tracking of the head. The system is full duplex with other systems allowing immersive collaboration. Lighting and sound generation take place close to the user's head. A haptic interface device allows the user to grab the display/sensor array and move it about. Motion acts as a planar selection device for 3D data. Planar force feedback allows a user to “feel” the data. Users see not only each other through display windows, but can also see the positions and orientations of each others' planar selections of shared 3D models or data.

Movable audio/video communication interface system

Provision of a user interface for computing devices equipped with a touchscreen display and a digital camera that magnifies a portion of a displayed image within a user's gaze. A user may calibrate their mobile device by touching a portion of the touchscreen with one or more fingers and following a moving image on the display with their eyes. The mobile device may track where a user is looking, and if the user is looking at the mobile device display, a portion of the display in the vicinity of the user's gaze may be enlarged. In an embodiment, if the user is looking at a virtual keyboard, key icons near the user's gaze may be increased in size commensurate with the user's finger tip size. In this manner, a user can accurately select individual keys in a virtual keyboard that fits within a mobile device display.

Apparatus and methods for dynamically correlating virtual keyboard dimensions to user finger size

The development and testing of two techniques of texture analysis based on different mathematical tools—the semivariogram and the Fourier spectra—are presented. These are also compared against a benchmark approach: the Gray-Level Co-occurrence Matrix. The three methods and their implementation are briefly described. Three series of experiments have been prepared to test the performance of these methods in various classification contexts. These contexts are simulated by varying the number, type and visual likeness of the texture patches used in classification tests. More specifically, their ability to correctly classify, separate, and associate texture patches is assessed. Results suggest that the classification context has an important impact on performance rates of all methods. The variogram-based and the Gray-Tone Dependency Matrix methods were generally superior, each one in particular contexts.

/pdf/comparing-texture-analysis-methods-through-classification-3ihzc4evj8.pdf

Comparing Texture Analysis Methods through Classification

Learning-based algorithm selection for image segmentation

In this paper, a face recognition algorithm using multi feature and Radial basis Function Network (RBFN) is proposed. The algorithm consists of three steps. In the first step, a coarse classification is performed using Fourier frequency spectrum feature, and only the first k gallery images with minimum Euclidean distance to the probe image are retained. In the second step, the Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) features of frequency spectrum are extracted, which will be taken as the input of the RBFN in the third step. In the last step, the classification is carried out by using RBFN. The proposed approach has been tested on ORL face database and Shimon database. The experimental results have demonstrated that the performance of this algorithm is much superior to the other algorithms on the same database.

Face recognition using multi-feature and radial basis function network

Using different algorithms to segment different images is a quite straightforward strategy for automated image segmentation. But the difficulty of the optimal algorithm selection has prevented it from being used for many years. In this paper, a framework of algorithm selection system is proposed to achieve automated image segmentation. Off-line learning scheme is adopted to make use of interactive segmentation evaluation. During training, both the performance ranks of candidate algorithms on every image and image features are used to train a predictor. Then, the performance ranks of all candidates will be predicted according to image features. Finally, the algorithm with the highest rank will be regarded as optimal and applied to the image. A simulation system is constructed to select optimal segmentation algorithm from four candidates for synthetic images. In this system, histogram is used as image feature, the number of misclassified pixels and computation expenses are used to facilitate interactive segmentation evaluation, and Principle Components Analysis (PCA) and Support Vector Machine (SVM) are used to construct the predictor. The system is tested on 9000 images by comparing with the manual selection. The best algorithms are selected for 84.90% of cases. If the second best algorithm is also regrded as acceptable, more than 97.5% of images can be properly segmented. The satisfied results demonstrate that this study has provided a promising approach to automated image segmentation.

Optimal selection of image segmentation algorithms based on performance prediction

With the development of the ubiquitous telephone network, speaker recognition-a more direct authentication technology attracts further attention. In this paper, an overview on the primary parts of current speaker recognition technology-modeling technology is presented. Based on the typical system structure we discussed the effective modeling techniques for both speaker and background, and distortion compensation techniques. Some issues on the technical challenges of current speaker recognition applications are also included so that some potential future trend in research, development and applications may be concluded.

An overview of modeling technology of speaker recognition

The IIR filter design under the mixed-criterion of H/sub 2/ norm and /spl infin/ norm is proposed in this paper, and genetic algorithms (GA) is introduced to realize the filter design based on such criterion. By this approach, the disadvantage of other conventional methods is avoided and the satisfied performance of the designed filter is obtained. The evaluating example shows that the filter designed by GA is superior to conventional Butterworth filter in either the optimization capability of design method or the performance of designed filter.

http://www.wseas.us/e-library/conferences/joint2002/451-233.pdf

Zhao Rongchun

Papers

Learning-based algorithm selection for image segmentation

Face recognition using multi-feature and radial basis function network

Optimal selection of image segmentation algorithms based on performance prediction

An overview of modeling technology of speaker recognition

Optimizing the design of IIR filter via genetic algorithm