Azizi Abdullah

Fixed partitioning and salient points with MPEG-7 cluster correlograms for image categorization

This paper compares fixed partitioning and salient points schemes for dividing an image into patches, in combination with low-level MPEG-7 visual descriptors to represent the patches with particular patterns. A clustering technique is applied to construct a compact representation by grouping similar patterns into a cluster codebook. The codebook will then be used to encode the patterns into visual keywords. In order to obtain high-level information about the relational context of an image, a correlogram is constructed from the spatial relations between visual keyword indices in an image. For classifying images a k-nearest neighbors (k-NN) and a support vector machine (SVM) algorithm are used and compared. The techniques are compared to other methods on two well-known datasets, namely Corel and PASCAL. To measure the performance of the proposed algorithms, average precision, a confusion matrix, and ROC-curves are used. The results show that the cluster correlogram outperforms the cluster histogram. The saliency based scheme performs similarly to the fixed partitioning scheme and the SVM significantly outperforms the k-NN classifier. Finally, we demonstrate the robustness to noise, photometric, and geometric distortions.

An Ensemble of Deep Support Vector Machines for Image Categorization

This paper presents the deep support vector machine (D-SVM) inspired by the increasing popularity of deep belief networks for image recognition. Our deep SVM trains an SVM in the standard way and then uses the kernel activations of support vectors as inputs for training another SVM at the next layer. In this way, instead of the normal linear combination of kernel activations, we can create non-linear combinations of kernel activations on prototype examples. Furthermore, we combine different descriptors in an ensemble of deep SVMs where the product rule is used for combining probability estimates of the different classifiers. We have performed experiments on 20 classes from the Caltech object database and 10 classes from the Corel dataset. The results show that our ensemble of deep SVMs significantly outperforms the naive approach that combines all descriptors directly in a very large single input vector for an SVM. Furthermore, our ensemble of D-SVMs achieves an accuracy of 95.2% on the Corel dataset with 10 classes, which is the best performance reported in literature until now.

Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study

Recent research in image recognition has shown that combining multiple descriptors is a very useful way to improve classification performance. Furthermore, the use of spatial pyramids that compute descriptors at multiple spatial resolution levels generally increases the discriminative power of the descriptors. In this paper we focus on combination methods that combine multiple descriptors at multiple spatial resolution levels. A possible problem of the naive solution to create one large input vector for a machine learning classifier such as a support vector machine, is that the input vector becomes of very large dimensionality, which can increase problems of overfitting and hinder generalization performance. Therefore we propose the use of stacking support vector machines where at the first layer each support vector machine receives the input constructed by each single descriptor and is trained to compute the right output class. A second layer support vector machine is then used to combine the class probabilities of all trained first layer support vector models to learn the right output class given these reduced input vectors. We have performed experiments on 20 classes from the Caltech object database with 10 different single descriptors at 3 different resolutions. The results show that our 2-layer stacking approach outperforms the naive approach that combines all descriptors directly in a very large single input vector.

Unimodal thresholding for Laplacian-based Canny-Deriche filter

A unimodal thresholding method for the Laplacian-based Canny-Deriche edge detector featuring a double-thresholding approach and reconstruction strategy was proposed. In this method, an improved image segmentation technique derived from an image histogram was developed. The accuracy of the segmentation was compared with the Otsu, Rosin, and Canny-hysteresis techniques. It was shown that the proposed method is more robust and accurate in detecting edges, resulting in a sensitivity of consistently more than 17.1%, with a standard deviation of less than 0.087, and a figure of merit (FOM) greater than 0.787 for all images tested in this study.

CIREC : Cluster Correlogram Image Retrieval and Categorization using MPEG-7 Descriptors

Content-based image retrieval is generally about understanding of information in the images concerned. The more the system is able to understand the content of images the more effective it will be in retrieving desired images. In this paper, we have developed a method that combines a clustering technique with the color correlation histogram. According to this method, an image is divided into several blocks using a fixed partitioning scheme. Then, selected lower-level MPEG-7 features are used to represent the partitions and a clustering technique is applied to group similar patterns into a cluster. The cluster indices will then be used to construct a corellogram which is not based on the lower-level primitives, but on regions consisting of a distribution of primitives. The result shows that the proposed method significantly outperforms other methods on image retrieval and categorization

A linear model based on Kalman filter for improving neural network classification performance

This paper proposes a method to improve neural network classification performance.A linear model was used as post processing of neural network.The parameters of linear model was estimated using Kalman filter iteration.The method can be applied to classify an object regardless of the type of feature.The method has been validated with five different datasets. Neural network has been applied in several classification problems such as in medical diagnosis, handwriting recognition, and product inspection, with a good classification performance. The performance of a neural network is characterized by the neural networks structure, transfer function, and learning algorithm. However, a neural network classifier tends to be weak if it uses an inappropriate structure. The neural networks structure depends on the complexity of the relationship between the input and the output. There are no exact rules that can be used to determine the neural networks structure. Therefore, studies in improving neural network classification performance without changing the neural networks structure is a challenging issue. This paper proposes a method to improve neural network classification performance by constructing a linear model based on the Kalman filter as a post processing. The linear model transforms the predicted output of the neural network to a value close to the desired output by using the linear combination of the object features and the predicted output. This simple transformation will reduce the error of neural network and improve classification performance. The Kalman filter iteration is used to estimate the parameters of the linear model. Five datasets from various domains with various characteristics, such as attribute types, the number of attributes, the number of samples, and the number of classes, were used for empirical validation. The validation results show that the linear model based on the Kalman filter can improve the performance of the original neural network.

Soft computing applications and intelligent systems: Second international multi-conference on artificial intelligence technology, m-cait 2013 shah alam, august 28-29, 2013 Proceedings

The determination of real world coordinate from image coordinate has many applications in computer vision. This paper proposes the algorithm for determination of real world coordinate of a point on a plane from its image coordinate using single calibrated camera based on simple analytic geometry. Experiment has been done using the image of chessboard pattern taken from five different views. The experiment result shows that exact real world coordinate and its approximation lie on the same plane and there are no significant difference between exact real world coordinate and its approximation.

Ensembles of novel visual keywords descriptors for image categorization

Object recognition systems need effective image descriptors to obtain good performance levels. Currently, the most widely used image descriptor is the SIFT descriptor that computes histograms of orientation gradients around points in an image. A possible problem of this approach is that the number of features becomes very large when a dense grid is used where the histograms are computed and combined for many different points. The current dominating solution to this problem is to use a clustering method to create a visual codebook that is exploited by an appearance based descriptor to create a histogram of visual keywords present in an image. In this paper we introduce several novel bag of visual keywords methods and compare them with the currently dominating hard bag-of-features (HBOF) approach that uses a hard assignment scheme to compute cluster frequencies. Furthermore, we combine all descriptors with a spatial pyramid and two ensemble classifiers. Experimental results on 10 and 101 classes of the Caltech-101 object database show that our novel methods significantly outperform the traditional HBOF approach and that our ensemble methods obtain state-of-the-art performance levels.

Arabic calligraphy identification for Digital Jawi Paleography using triangle blocks

Digital Jawi Paleography is a field of research that helps paleographers to identify authors, origin and date of Jawi manuscripts. This research is important because of the existence of a huge amount of Malay manuscripts with unidentified authors, origin and date. Most researches in the area are for Roman and Hebrew text, whereas researches for Jawi text have just begun recently. In this paper, a novel technique is proposed in order to identify types of Arabic calligraphy in Malay ancient manuscripts that were written in Jawi. The novel technique is based on the triangle blocks that were adapted from scalene triangle. Twenty-one features have been extracted from the triangle blocks.

A robust crack detection method for non-uniform distributions of coloured and textured image

A robust automatic crack detection method for nonuniform colour distributions on texture images is proposed. In this method a new image segmentation technique is developed while the Hough transform is used for feature extraction. Meanwhile, the detection is based on standard discriminant analysis, featuring Wilks λ selection criteria. The methods and procedures were tested on commercial biscuit crackers, resulting in the specificity and sensitivity of more than 98% and 88%, respectively. Since the algorithm is implemented in software, the system could be programmed to inspect other manufactured products.