International Journal of Pattern Recognition and Artificial Intelligence

A first attempt to incorporate Fuzzy Cognitive Maps (FCMs), in pattern classification applications is performed in this paper. Fuzzy Cognitive Maps, as an illustrative causative representation of modeling and manipulation of complex systems, can be used to model the behavior of any system. By transforming a pattern classification problem into a problem of discovering the way the sets of patterns interact with each other and with the classes that they belong to, we could describe the problem in terms of Fuzzy Cognitive Maps. More precisely, some FCM architectures are introduced and studied with respect to their pattern recognition abilities. An efficient novel hybrid classifier is proposed as an alternative classification structure, which exploits both neural networks and FCMs to ensure improved classification capabilities. Appropriate experiments with four well-known benchmark classification problems and a typical computer vision application establish the usefulness of the Fuzzy Cognitive Maps, in a pattern recognition research field. Moreover, the present paper introduces the use of more flexible FCMs by incorporating nodes with adaptively adjusted activation functions. This advanced feature gives more degrees of freedom in the FCM structure to learn and store knowledge, as needed in pattern recognition tasks.

In this paper we present an improved support vector machines (SVMs) watermarking system for still images and video sequences. By a thorough study on feature selection for training SVM, the proposed system shows significant improvements on computation efficiency and robustness to various attacks. The improved algorithm is extended to be a scene-based video watermarking technique. In a given scene, the algorithm uses the first h' frames to train an embedding SVM, and uses the trained SVM to watermark the rest of the frames. In the extracting phrase, the detector uses only the center h frames of the first h' frames to train an extracting SVM. The final extracted watermark in a given scene is the average of watermarks extracted from the remaining frames. Watermarks are embedded in l longest scenes of a video such that it is computationally efficient and capable to resist possible frames swapping/deleting/duplicating attacks. Two collusion attacks, namely temporal frame averaging and watermark estimation remodulation, on video watermarking are discussed and examined. The proposed video watermarking algorithm is shown to be robust to compression and collusion attacks, and it is novel and practical for SVM-applications.

There are several commonly accepted clustering quality measures (clustering quality as opposed to cluster quality) such as the rand index, the adjusted rand index and the jacquard index. Each of these however is based on comparing the partition produced by the clustering process to a correct partition. They can therefore only be used to determine the quality of a clustering process when the correct partition is known. This paper therefore proposes another clustering quality measure that does not require the comparison to a correct partition.

The proposed metric is based on the assumption that the proximities between the membership vectors should correlate positively with the proximities between the objects which may be the proximities between their feature vectors. The values of the components of the membership vector, corresponding to a pattern, are the membership degrees of the pattern in the various clusters. The membership vector is just another object data vector or type of feature vector with the feature values for an object being the membership values of the object in the various clusters. Based on this premise, this paper describes some new cluster quality metrics derived from standard correlation measures and other proposed correlation metrics.

Simulations on data with a wide range of clusterability or separability show that the approach of comparing the proximity matrix based on the membership matrix to the object proximity matrix is quite effective.

We present an efficient algorithm for computing a sub-optimal Earth Movers' Distance (EMD) between multidimensional histograms called EMD-g_f, which is not limited to any type of measurement. Some algorithms that find a cross-bin distance between histograms have been proposed in the literature. Nevertheless, most of this research has been applied on 1D-histograms or on nD-histograms but with limited types of measurements. The EMD is a cross-bin distance between nD-histograms with any ground distance. Experimental validation shows that it obtains good retrieval results although the main drawback of this method is its cubic computational cost, O(z³), z being the total number of bins. The worst-case complexity of EMD-g_f is O(z²), although the obtained average computational cost in the experiments is near O(m²), where m represents the number of bins per dimension, which is clearly lower than the computational cost of the EMD algorithm. Moreover, the experiments using real data show similar retrieval results.

There is no parametric formulation of corner, so the conventional Hough transform cannot be employed to detect corners directly. A random corner detection method is developed in this paper based on a new concept "accumulative intersection space" under Monte Carlo scheme. This method transforms the corner detection in the image space into local maxima localization in the accumulative intersection space where the intersections are accumulated by random computations. The proposed algorithm has been demonstrated by both theory and experiments. The proposed algorithm is isotropic, robust to image rotation, insensitive to noise and false corners on diagonal edges. Unlike the other existing contour based corner detection methods, our algorithm can effectively avoid the influence of the edge detectors, such as rounding corners or line interceptions. Extensive comparisons among our approach and the other detectors including Harris operator, Fei Shen and Han Wang detector, Han Wang and Brady detector, Foveated Visual Search method and SIFT feature, have shown the effectiveness of our method.

Classification of gene expression samples is a core task in microarray data analysis. How to reduce thousands of genes and to select a suitable classifier are two key issues for gene expression data classification. This paper introduces a framework on combining both feature extraction and classifier simultaneously. Considering the non-negativity, high dimensionality and small sample size, we apply a discriminative mixture model which is designed for non-negative gene express data classification via non-negative matrix factorization (NMF) for dimension reduction. In order to enhance the sparseness of training data for fast learning of the mixture model, a generalized NMF is also adopted. Experimental results on several real gene expression datasets show that the classification accuracy, stability and decision quality can be significantly improved by using the generalized method, and the proposed method can give better performance than some previous reported results on the same datasets.

An improved scheme for integrity protection of binary images representing text documents based on the topologies of these images is proposed. The image skeleton and the inverse skeleton are found through thinning and the skeleton signature is combined with watermark information. The result is encrypted asymmetrically and hidden in embeddable locations of the image. A series of attack experiments are conducted to demonstrate that the approach is capable of detecting tampering. Even single malicious pixel modifications are detected. The approach has a lower computation cost than previous methods.

In recent years stochastic context-free grammars have been shown to be effective in modeling human activities because of the hierarchical structures they represent. However, most of the research in this area has yet to address the issue of learning the activity grammars from a noisy input source, namely, video. In this paper, we present a framework for identifying noise and recovering the basic activity grammar from a noisy symbol string produced by video. We identify the noise symbols by finding the set of non-noise symbols that optimally compresses the training data, where the optimality of compression is measured using an MDL criterion. We show the robustness of our system to noise and its effectiveness in learning the basic structure of human activity, through experiments with artificial data and a real video sequence from a local convenience store.

Some facial features that differ from an ordinary face should be identified by a computer when generating a facial caricature. These distinctive facial features are called self-features. Compared with traditional Mean Face Model (MFM) that is unable to quantify these self-features well, a Self-Reference Model (SRM) is presented in this paper. Firstly, based on the physiology structure of a front face, a self-reference is found, and this reference is used to measure the self-features. According to the self-reference, some standard facial parameters are worked out by collecting statistic data of many facial images. Then, in an input face image, by evaluating some differences between the input face and the standard facial parameters, the self-features are properly estimated and quantified. Finally, by analyzing some caricatures produced by caricaturists, the SRM can prove the validity of the proposed Algorithm.