Unlike the time-consuming contour tracking method of snakes [5] which requires a considerable number of iterated computations before contours are successfully tracked down, we present a faster and accurate model-based landmarks" tracking method where a single iteration of the dynamic programming is sufficient to obtain a local minimum to an integral measure of the elastic and the image energy functionals. The key lies in choosing a relatively small number of salient land-marks", or features of objects, rather than their contours as a target of tracking within the image structure. The landmarks comprising singular points along the model contours are tracked down within the image structure all inside restricted search areas of 41 41 pixels whose respective locations in image structure are dictated by their locations in the model. A Manhattan distance and a template corner detection function of Singh and Shneier [7] are used as elastic energy and image energy respectively in the algorithm. A first approximation to the image contour is obtained in our method by applying the thin-plate spline transformation of Bookstein [2] using these landmarks as fixed points of the transformation which is capable of preserving a global shape information of the model including the relative configuration of landmarks and consequently surrounding contours of the model in the image structure. The actual image contours are further tracked down by applying an active edge tracker using now simplified line search segments so that individual differences persisting between the mapped model contour are substantially eliminated. We have applied our method tentatively to portraits of a class album to demonstrate the effectiveness of the method. Our experiments convincingly show that using only about 11 feature points our method provides not only a much improved computational complexity requiring only 0.94sec. in CPU time by SGI's indigo2 but also more accurate shape representations than those obtained by the snakes methods. The method is powerful in a problem domain where the model-based approach is applicable, possibly allowing real time processing because a most time consuming algorithm of corner template evaluation can be easily implemented by parallel processing firmware.

One fundamental issue in multiagent reinforcement learning is how to deal with the limited local knowledge of an agent in order to achieve effective learning. In this paper, we argue that this issue can be more effectively solved if agents are equipped with a consistent global view. We achieve this by requiring agents to follow an interacting protocol. The properties of the protocol are derived and theoretically analyzed. A distributed protocol that satisfies these properties is presented. The experimental evaluations are conducted for a well-known test-case (i.e., pursuit game) in the context of two learning algorithms. The results show that the protocol is effective and the reinforcement learning algorithms using it perform much better.

A class of multimodulus algorithms (MMA(p)) optimized by an optimal step-size (OS) for blind equalization are firstly investigated in this letter. The multimodulus (MM) criterion is essentially a split cost function that separately implements the real and imaginary part of the signal, hence the phase can be recovered jointly with equalization. More importantly, the step-size leading to the minimum of the MM criterion along the search direction can be obtained algebraically among the roots of a higher-order polynomial at each iteration, thus a robust optimal step-size multimodulus algorithm (OS-MMA(p)) is developed. Experimental results demonstrate improved performance of the proposed algorithm in mitigating the inter-symbol interference (ISI) compared with the OS constant modulus algorithm (OS-CMA). Besides, the computational complexity can be reduced by the proposed OS-MMA(2) algorithm.

Minutiae pattern extraction plays a crucial role in fingerprint registration and identification for electronic applications. However, the extraction accuracy is seriously compromised by the presence of contaminated ridge lines and complex background scenarios. General image processing-based methods, which rely on many prior hypotheses, fail to effectively handle minutiae extraction in complex scenarios. Previous works have shown that CNN-based methods can perform well in object detection tasks. However, the deep neural networks (DNNs)-based methods are restricted by the limitation of public labeled datasets due to legitimate privacy concerns. To address these challenges comprehensively, this paper presents a fully automated minutiae extraction method leveraging DNNs. Firstly, we create a fingerprint minutiae dataset using a semi-automated minutiae annotation algorithm. Subsequently, we propose a minutiae extraction model based on Residual Networks (Resnet) that enables end-to-end prediction of minutiae. Moreover, we introduce a novel non-maximal suppression (NMS) procedure, guided by the Generalized Intersection over Union (GIoU) metric, during the inference phase to effectively handle outliers. Experimental evaluations conducted on the NIST SD4 and FVC 2004 databases demonstrate the superiority of the proposed method over existing state-of-the-art minutiae extraction approaches.

Considering the inaccuracy of image registration, we propose a new regularization restoration algorithm to solve the ill-posed super-resolution (SR) problem. Registration error is used to obtain cross-channel error information caused by inaccurate image registration. The registration error is considered as the noise mean added into the within-channel observation noise which is known as Additive White Gaussian Noise (AWGN). Based on this consideration, two constraints are regulated pixel by pixel within the framework of Miller's regularization. Regularization parameters connect the two constraints to construct a cost function. The regularization parameters are estimated adaptively in each pixel in terms of the registration error and in each observation channel in terms of the AWGN. In the iterative implementation of the proposed algorithm, sub-sampling operation and sampling aliasing in the detector model are dealt with respectively to make the restored HR image approach the original one further. The transpose of the sub-sampling operation is implemented by nearest interpolation. Simulations show that the proposed regularization algorithm can restore HR images with much sharper edges and greater SNR improvement.

This paper studies using price incentives to shift bandwidth demand from peak to non-peak periods. In particular, cost discounts decrease as peak monthly usage increases. We take into account the delay sensitivity of different apps: during peak hours, the usage of hard real-time applications (HRAS) is not counted in the user's monthly data cap, while the usage of other applications (OAS) is counted in the user's monthly data cap. As a result, users may voluntarily delay or abandon OAS in order to get a higher fee discount. Then, a new data rate control algorithm is proposed. The algorithm allocates the data rate according to the priority of the source, which is determined by two factors: (I) the allocated data rate; and (II) the waiting time.

This paper describes a new system for extracting and classifying bibliography regions from the color image of a book cover. The same as all the color image processing, the segmentation of color space is an essential and important step in our system; and here HSI color space is adopted rather than RGB color space. The color space is segmented into achromatic and chromatic regions first; and the segmentation is completed after thresholding the intensity histogram of the achromatic region and the hue histogram of the chromatic region. Then text region extraction and classification follows. After detecting fundamental features (stroke width and local label width) text regions are determined by comparing smeared blocks to the original candidate image. Based on the general cover design model, text regions are classified into author region, title region, and publisher region furthermore, and a bibliography image is obtained as a result, without applying OCR. The appearance of the book is 3D reconstructed as well. In this paper, two examples are presented.

Traffic classification has recently gained much attention in both academic and industrial research communities. Many machine learning methods have been proposed to tackle this problem and have shown good results. However, when applied to traffic with out-of-sequence packets, the accuracy of existing machine learning approaches decreases dramatically. We observe the main reason is that the out-of-sequence packets change the spatial representation of feature vectors, which means the property of linear mapping relation among features used in machine learning approaches cannot hold any more. To address this problem, this paper proposes an Improved Dynamic Time Warping (IDTW) method, which can align two feature vectors using non-linear alignment. Experimental results on two real traces show that IDTW achieves better classification accuracy in out-of-sequence traffic classification, in comparison to existing machine learning approaches.