Kwok-Ping Chan

Real-time large scale near-duplicate web video retrieval

Near-duplicate video retrieval is becoming more and more important with the exponential growth of the Web. Though various approaches have been proposed to address this problem, they are mainly focusing on the retrieval accuracy while infeasible to query on Web scale video database in real time. This paper proposes a novel method to address the efficiency and scalability issues for near-duplicate We video retrieval. We introduce a compact spatiotemporal feature to represent videos and construct an efficient data structure to index the feature to achieve real-time retrieving performance. This novel feature leverages relative gray-level intensity distribution within a frame and temporal structure of videos along frame sequence. The new index structure is proposed based on inverted file to allow for fast histogram intersection computation between videos. To demonstrate the effectiveness and efficiency of the proposed methods we evaluate its performance on an open Web video data set containing about 10K videos and compare it with four existing methods in terms of precision and time complexity. We also test our method on a data set containing about 50K videos and 11M key-frames. It takes on average 17ms to execute a query against the whole 50K Web video data set.

Fuzzy-attribute graph with application to Chinese character recognition

To include fuzzy properties in solving some types of problems, the attributes graph is extended to provide a fuzzy-attribute graph (FAG). With such an extension, equality of attributes can no longer be used when matching of FAGs is considered, as equality of two fuzzy sets is too stringent a condition. A measure for matching two FAGs is suggested. The measure has its interpretation in fuzzy logic. The result of applying the model to the recognition of handprinted Chinese characters is presented. >

Restricted Random Testing

This paper presents a novel adaptation of traditional random testing, called Restricted Random Testing (RRT). RRT offers a significant improvement over random testing, as measured by the F-measure. This paper describes the ideology behind RRT and explains its algorithm. RRTs performance is examined using several experiments, the results of which are presented and discussed.

RESTRICTED RANDOM TESTING: ADAPTIVE RANDOM TESTING BY EXCLUSION

Restricted Random Testing (RRT) is a new method of testing software that improves upon traditional Random Testing (RT) techniques. Research has indicated that failure patterns (portions of an input domain which, when executed, cause the program to fail or reveal an error) can influence the effectiveness of testing strategies. For certain types of failure patterns, it has been found that a widespread and even distribution of test cases in the input domain can be significantly more effective at detecting failure compared with ordinary RT. Testing methods based on RT, but which aim to achieve even and widespread distributions, have been called Adaptive Random Testing (ART) strategies. One implementation of ART is RRT. RRT uses exclusion zones around executed, but non-failure-causing, test cases to restrict the regions of the input domain from which subsequent test cases may be drawn. In this paper, we introduce the motivation behind RRT, explain the algorithm and detail some empirical analyses carried out to examine the effectiveness of the method. Two versions of RRT are presented: Ordinary RRT (ORRT) and Normalized RRT (NRRT). The two versions share the same fundamental algorithm, but differ in their treatment of non-homogeneous input domains. Investigations into the use of alternative exclusion shapes are outlined, and a simple technique for reducing the computational overheads of RRT, prompted by the alternative exclusion shape investigations, is also explained. The performance of RRT is compared with RT and another ART method based on maximized minimum test case separation (DART), showing excellent improvement over RT and a very favorable comparison with DART.

Nonparametric discriminant HMM and application to facial expression recognition

This paper presents a nonparametric discriminant HMM and applies it to facial expression recognition. In the proposed HMM, we introduce an effective nonparametric output probability estimation method to increase the discrimination ability at both hidden state level and class level. The proposed method uses a nonparametric adaptive kernel to utilize information from all classes and improve the discrimination at class level. The discrimination between hidden states is increased by defining membership coefficients which associate each reference vector with hidden states. The adaption of such coefficients is obtained by the expectation maximization (EM) method. Furthermore, we present a general formula for the estimation of output probability, which provides a way to develop new HMMs. Finally, we evaluate the performance of the proposed method on the CMU expression database and compare it with other nonparametric HMMs.

Symmetric extension methods for M-channel linear-phase perfect-reconstruction filter banks

The symmetric extension method has been shown to he an efficient way for subband processing of finite-length sequences. This paper presents an extension of this method to general linear-phase perfect-reconstruction filter banks. We derive constraints on the length and symmetry polarity of the permissible filter banks and propose a new design algorithm. In the algorithm, different symmetric sequences are formulated in a unified form based on the circular-symmetry framework. The length constraints in symmetrically extending the input sequence and windowing the subband sequences are investigated. The effect of shifting the input sequence is included. When the algorithm is applied to equal-length filter banks, we explicitly show that symmetric extension methods can always be constructed to replace the circular convolution approach.

Clustering of clusters

Abstract An algorithm for the clustering of existing clusters is introduced in this paper. The algorithm was adopted from fuzzy-c-mean and modifications made to take into account the extra information, i.e. some data samples already form clusters. Partition coefficients, together with some other criteria, are used for testing cluster validity. The method was applied on Chinese character recognition and an encouraging result was obtained.

A revisit of adaptive random testing by restriction

Adaptive random testing is a black box testing method based on the intuition that random testing failure-finding efficiency can be improved upon, in certain situations, by ensuring a more widespread and evenly distributed spread of test cases in the input domain. One way of achieving this distribution is through the use of exclusion zones and restriction, resulting in a method called restricted random testing (RRT). Recent investigations into the RRT method have revealed several interesting and significant insights. A method of reducing the computational overheads of testing methods by partitioning an input domain, and applying the method to only one of the subdomains, mapping the test cases to other subdomains, has recently been introduced. This method, called mirroring, in addition to alleviating computational costs, has some properties which fit nicely with the insights into RRT, offering solutions to some possible shortcomings of RRT. In this paper we discuss the RRT method and additional insights; we explain mirroring; and we detail applications of mirroring to RRT. The mirror RRT method proves to be a very attractive variation of RRT

Normalized restricted random testing

Restricted Random Testing (RRT) is a new method of testing software that improves upon traditional random testing (RT) techniques. This paper presents new data in support of the efficiency of RRT, and presents a variation of the algorithm, Normalized Restricted Random Testing (NRRT). NRRT permits the tester to have better information about the target exclusion rate (R) of RRT, the main control parameter of the method. We examine the performance of the NRRT and Original RRT (ORRT) methods using simulations and experiments, and offer some guidance for their use in practice.

Evolutionary Cross-Domain Discriminative Hessian Eigenmaps

Is it possible to train a learning model to separate tigers from elks when we have 1) labeled samples of leopard and zebra and 2) unlabelled samples of tiger and elk at hand? Cross-domain learning algorithms can be used to solve the above problem. However, existing cross-domain algorithms cannot be applied for dimension reduction, which plays a key role in computer vision tasks, e.g., face recognition and web image annotation. This paper envisions the cross-domain discriminative dimension reduction to provide an effective solution for cross-domain dimension reduction. In particular, we propose the cross-domain discriminative Hessian Eigenmaps or CDHE for short. CDHE connects training and test samples by minimizing the quadratic distance between the distribution of the training set and that of the test set. Therefore, a common subspace for data representation can be well preserved. Furthermore, we basically expect the discriminative information used to separate leopards and zebra can be shared to separate tigers and elks, and thus we have a chance to duly address the above question. Margin maximization principle is adopted in CDHE so the discriminative information for separating different classes (e.g., leopard and zebra here) can be well preserved. Finally, CDHE encodes the local geometry of each training class (e.g., leopard and zebra here) in the local tangent space which is locally isometric to the data manifold and thus CDHE preserves the intraclass local geometry. The objective function of CDHE is not convex, so the gradient descent strategy can only find a local optimal solution. In this paper, we carefully design an evolutionary search strategy to find a better solution of CDHE. Experimental evidence on both synthetic and real word image datasets demonstrates the effectiveness of CDHE for cross-domain web image annotation and face recognition.