Image Feature Space Research Articles

Correlation coefficient based supervised locally linear embedding for pulmonary nodule recognition

Background and objectiveDimensionality reduction techniques are developed to suppress the negative effects of high dimensional feature space of lung CT images on classification performance in computer aided detection (CAD) systems for pulmonary nodule detection. MethodsAn improved supervised locally linear embedding (SLLE) algorithm is proposed based on the concept of correlation coefficient. The Spearman's rank correlation coefficient is introduced to adjust the distance metric in the SLLE algorithm to ensure that more suitable neighborhood points could be identified, and thus to enhance the discriminating power of embedded data. The proposed Spearman's rank correlation coefficient based SLLE (SC2SLLE) is implemented and validated in our pilot CAD system using a clinical dataset collected from the publicly available lung image database consortium and image database resource initiative (LICD-IDRI). Particularly, a representative CAD system for solitary pulmonary nodule detection is designed and implemented. After a sequential medical image processing steps, 64 nodules and 140 non-nodules are extracted, and 34 representative features are calculated. The SC2SLLE, as well as SLLE and LLE algorithm, are applied to reduce the dimensionality. Several quantitative measurements are also used to evaluate and compare the performances. ResultsUsing a 5-fold cross-validation methodology, the proposed algorithm achieves 87.65% accuracy, 79.23% sensitivity, 91.43% specificity, and 8.57% false positive rate, on average. Experimental results indicate that the proposed algorithm outperforms the original locally linear embedding and SLLE coupled with the support vector machine (SVM) classifier. ConclusionsBased on the preliminary results from a limited number of nodules in our dataset, this study demonstrates the great potential to improve the performance of a CAD system for nodule detection using the proposed SC2SLLE.

Development of a calibrating algorithm for Delta Robot’s visual positioning based on artificial neural network

Delta robot with vision system can automatically control the end-actuator to accurately grasp moving objects on the conveyor belt. Establishment of the mapping relationship between the image feature space and the robot working space form a closed-loop chain for transformational link between the robot coordinate, camera coordinate and conveyor belt coordinate. The vision system calibration is a basic problem of robot vision research and implementation. The artificial neural networks (ANN) which has learning ability, adaptive ability and nonlinear function approximation ability can establish the nonlinear relationship between space points and pixel points to complete accurate calibration of the vision system. The convergence speed of calibration algorithm affects the real-time visual servo system. The calibration precision, generalization ability and calibration space of algorithm influence the robot grasping accuracy. Therefore, a new calibration technique for delta robot’s vision system was presented in this paper. The algorithm combines ANN with Faugeras vision system calibration technology. The setting of the initial value, network structure and the choice of the activation function is based on the model of Faugeras vision system calibration algorithm, which makes the actual output of the network closer to the target output. Experiments proved that this algorithm has higher calibration accuracy and generalization ability compared with the conventional calibration algorithm, as well as faster convergence speed compared with the conventional artificial neural network structure in the case of high calibration accuracy.

Real-time object tracking via compressive feature selection

Recently, compressive tracking (CT) has been widely proposed for its efficiency, accuracy and robustness on many challenging sequences. Its appearance model employs non-adaptive random projections that preserve the structure of the image feature space. A very sparse measurement matrix is used to extract features by multiplying it with the feature vector of the image patch. An adaptive Bayes classifier is trained using both positive samples and negative samples to separate the target from background. On the CT framework, however, some features used for classification have weak discriminative abilities, which reduces the accuracy of the strong classifier. In this paper, we present an online compressive feature selection algorithm(CFS) based on the CT framework. It selects the features which have the largest margin when using them to classify positive samples and negative samples. For features that are not selected, we define a random learning rate to update them slowly. It makes those weak classifiers preserve more target information, which relieves the drift when the appearance of the target changes heavily. Therefore, the classifier trained with those discriminative features couples its score in many challenging sequences, which leads to a more robust tracker. Numerous experiments show that our tracker could achieve superior result beyond many state-of-the-art trackers.

Extended compressed tracking via random projection based on MSERs and online LS-SVM learning

The compressed tracking algorithm (CT tracker) is a well-known visual tracking method that models a target object׳s appearance through sparse random projection. However, the tracking results are not stable and robust due to the randomness of random projection. To solve this problem, a more stable and robust approach is proposed for visual tracking based on maximally stable extremal regions (MSERs), sparse random projection and online least squares SVM classifier (LS-SVM) learning. To obtain a relatively stable appearance model, the stable connected components of an object based on MSERs in image feature space are extracted. With the fusion of MSERs and sparse random projection, we model adaptive object appearance to adapt the variation of appearance. Additionally, an online closed-form LS-SVM is employed to quickly and robustly predict the target object location in a tracking by detection framework. Experimental results on benchmark sequences show the stability and robustness of the proposed algorithm compared with the existing CT-based trackers and other state-of-the-art trackers.

POSE-VIWEPOINT ADAPTIVE OBJECT TRACKING VIA ONLINE LEARNING APPROACH

In this paper, we propose an effective tracking algorithm with an appearance model based on features extracted from a video frame with posture variation and camera view point adaptation by employing the non-adaptive random projections that preserve the structure of the image feature space of objects. The existing online tracking algorithms update models with features from recent video frames and the numerous issues remain to be addressed despite on the improvement in tracking. The data-dependent adaptive appearance models often encounter the drift problems because the online algorithms does not get the required amount of data for online learning. So, we propose an effective tracking algorithm with an appearance model based on features extracted from a video frame.

Knowledge-Encoded Granular Neural Networks for Hyperspectral Remote Sensing Image Classification

Development of intelligent decision-making systems for complex problems, such as land covers classification of hyperspectral remote sensing (HSRS) images, requires efficient interpretation of available information through conceptual rather than numerical level. Granular neural network (GNN) in combination with the granular representation of information using linguistic terms is one such system. GNN takes the fuzzified input information and processes them with neural network (NN) architecture, where the network structure is transparent enough to interpret the processing steps. Further, knowledge encoding has been considered as one of the principal elements of intelligent decision-making systems. This paper proposes a new model of knowledge-encoded GNNs for land cover classification of HSRS images. Knowledge encoding is done using neighborhood rough sets (NRSs) that explore the local/contextual information. The encoded knowledge using NRS is obtained in the form of dependency rules with respect to the output class labels of land covers and these rules determine appropriate number of hidden nodes of GNNs. The dependency factors obtained during rule generation are used for initializing the connecting weights of GNNs. NRS is also used here in the selection of a subset of features for reducing the burden of high-dimensional fuzzy-granulated feature space of HSRS image. The proposed model thus explores jointly the advantages of fuzzy granulation, GNNs, and feature selection and knowledge encoding using NRS. Superiority of the model to similar other methods are justified in land covers classification of two HSRS images acquired by different remotely placed sensors.

A multi-scale method for urban tree canopy clustering recognition using high-resolution image

In this paper a constraint mean shift method is proposed for extracting urban tree canopy with the use of high-resolution image. Through establishing multi-scale pyramid image feature space by wavelet, features between layers can be combined as the constrained route for mean shift to realize self-adaptive decomposition and scale transfer in multi-scale feature space, then differences of internal and external structure in urban tree canopy and differences of average spectral radiant intensity are used as multi-scale feature space of wavelet to realize the preliminary clustering segmentation, finally we apply the supervised segmentation to extract tree canopy based on clustering feature. Experiments demonstrate that the proposed method can eliminate the over-detailed effect of image accurate extraction of the urban tree canopy can be achieved.

A Novel Active Learning Method in Relevance Feedback for Content-Based Remote Sensing Image Retrieval

Conventional relevance feedback (RF) schemes improve the performance of content-based image retrieval (CBIR) requiring the user to annotate a large number of images. To reduce the labeling effort of the user, this paper presents a novel active learning (AL) method to drive RF for retrieving remote sensing images from large archives in the framework of the support vector machine classifier. The proposed AL method is specifically designed for CBIR and defines an effective and as small as possible set of relevant and irrelevant images with regard to a general query image by jointly evaluating three criteria: uncertainty; diversity; and density of images in the archive. The uncertainty and diversity criteria aim at selecting the most informative images in the archive, whereas the density criterion goal is to choose the images that are representative of the underlying distribution of data in the archive. The proposed AL method assesses jointly the three criteria based on two successive steps. In the first step, the most uncertain (i.e., ambiguous) images are selected from the archive on the basis of the margin sampling strategy. In the second step, the images that are both diverse (i.e., distant) to each other and associated to the high-density regions of the image feature space in the archive are chosen from the most uncertain images. This step is achieved by a novel clustering-based strategy. The proposed AL method for driving the RF contributes to mitigate problems of unbalanced and biased set of relevant and irrelevant images. Experimental results show the effectiveness of the proposed AL method.

Learning multiple linear mappings for efficient single image super-resolution.

Example learning-based superresolution (SR) algorithms show promise for restoring a high-resolution (HR) image from a single low-resolution (LR) input. The most popular approaches, however, are either time- or space-intensive, which limits their practical applications in many resource-limited settings. In this paper, we propose a novel computationally efficient single image SR method that learns multiple linear mappings (MLM) to directly transform LR feature subspaces into HR subspaces. In particular, we first partition the large nonlinear feature space of LR images into a cluster of linear subspaces. Multiple LR subdictionaries are then learned, followed by inferring the corresponding HR subdictionaries based on the assumption that the LR-HR features share the same representation coefficients. We establish MLM from the input LR features to the desired HR outputs in order to achieve fast yet stable SR recovery. Furthermore, in order to suppress displeasing artifacts generated by the MLM-based method, we apply a fast nonlocal means algorithm to construct a simple yet effective similarity-based regularization term for SR enhancement. Experimental results indicate that our approach is both quantitatively and qualitatively superior to other application-oriented SR methods, while maintaining relatively low time and space complexity.

Adaptive weighted real‐time compressive tracking

Many tracking methods often suffer from the drift problems caused by appearance change. Therefore developing a robust online tracker is still a challenging test. Recently, a simple yet effective and efficient tracking algorithm has been proposed by compressive tracking (CT) paradigm to alleviate the drift to some degree. The CT tracker introduced an appearance model based on features extracted from the multi‐scale image feature space in the compressed domain. However, the CT tracker may detect the positive sample that is less important because it does not discriminatively consider the sample importance in its learning procedure. In this study, the authors integrate the sample importance into the CT tracker online learning procedure. They also add an efficient feature select method which can choose the most discriminative power weak classifier and employ the co‐training criterion into CT tracker to improve the tracking performance. Experiments show that the proposed tracker demonstrates the superior performance in robustness and efficiency than other state‐of‐the‐art trackers.

Fast Compressive Tracking.

It is a challenging task to develop effective and efficient appearance models for robust object tracking due to factors such as pose variation, illumination change, occlusion, and motion blur. Existing online tracking algorithms often update models with samples from observations in recent frames. Despite much success has been demonstrated, numerous issues remain to be addressed. First, while these adaptive appearance models are data-dependent, there does not exist sufficient amount of data for online algorithms to learn at the outset. Second, online tracking algorithms often encounter the drift problems. As a result of self-taught learning, misaligned samples are likely to be added and degrade the appearance models. In this paper, we propose a simple yet effective and efficient tracking algorithm with an appearance model based on features extracted from a multiscale image feature space with data-independent basis. The proposed appearance model employs non-adaptive random projections that preserve the structure of the image feature space of objects. A very sparse measurement matrix is constructed to efficiently extract the features for the appearance model. We compress sample images of the foreground target and the background using the same sparse measurement matrix. The tracking task is formulated as a binary classification via a naive Bayes classifier with online update in the compressed domain. A coarse-to-fine search strategy is adopted to further reduce the computational complexity in the detection procedure. The proposed compressive tracking algorithm runs in real-time and performs favorably against state-of-the-art methods on challenging sequences in terms of efficiency, accuracy and robustness.

ZASTOSOWANIE METODY HELLWIGA DO REDUKCJI WYMIARU PRZESTRZENI CECH OBRAZÓW USG TARCZYCY

This paper presents the use of Hellwig’s method for dimension reduction in feature space of thyroid ultrasound images. On the base of this method, the combination of three features with the greatest value of Hellwig’s index information capacity from the input set of 283 features was obtained. This set was used to build and test the classifiers. Classification results were compared with the results obtained for a set of 48 features obtained using correlation method. It turned out that the accuracy of classifiers built on the base of 3 features is not worse than the accuracy of classifiers built on the base of 48 features, and in some cases it is even higher. This suggests that the Hellwig’s method can be used as an effective method for dimension reduction in feature space for the future thyroid ultrasound images classification.

Combining data mining algorithm and object-based image analysis for detailed urban mapping of hyperspectral images

Image classification of roofing types, road pavements, and natural features can assist land-cover maps in further examining the effects of such features on health, pollution, and the microclimate in urban settings. Airborne hyperspectral sensors with high spectral and spatial resolutions can be employed for detailed characterization of urban areas. This study aims to develop a procedure that is instrumental for automated knowledge discovery and mapping of urban surface materials from a large feature space of hyperspectral images. Two different images over Universiti Putra Malaysia (UPM) and Kuala Lumpur (KL), Malaysia, were captured by using hyperspectral sensors with 20 and 128 bands. The images were used to explore the combined performance of a data mining (DM) algorithm and object-based image analysis (OBIA). A large number of attributes were discovered with the C4.5 DM algorithm, which also generated the classification model as a decision tree. The UPM and KL classified images achieved 93.42 and 88.36% overall accuracy. The high accuracy of object-based classification can be linked to the knowledge discovery produced by the DM algorithm. This algorithm increased the productivity of OBIA, expedited the process of attribute selection, and resulted in an easy-to-use representation of a knowledge model from a decision tree structure.

A novel CBIR system with WLLTSA and ULRGA

At present, relevance feedback (RF) has been widely applied in content-based image retrieval (CBIR) system. Local Regression and Global Alignment (LRGA) is a novel ranking algorithm used in CBIR system which utilizes RF technique. However, there are some problems in LRGA: (1) for handling the problem of out-of-sample, dimension reduction is used after RF, but it is time-consuming; (2) feature space of images is often assumed to be linear. While, classical manifold learning methods are sensitive to the Gaussian bandwidth parameter of Laplacian matrix and cannot be combined with RF either. To address problems above, this paper proposes a novel CBIR system. Firstly, we calculate the local curvature parameter of manifold utilizing the angle information in subspace to avoid local high curvature problem and then we propose a Warp Linear Local Tangent Space Alignment (WLLTSA) algorithm; furthermore, we propose a U-Local Regression and Global Alignment (ULRGA) ranking algorithm to rank low-dimensional image features. Curvature parameter is used in both WLLTSA and ULRGA to enhance robustness. A large amount of experimental results demonstrate the efficiency of our CBIR system.

A novel image retrieval method based on hybrid information descriptors

In this paper, we propose a novel image retrieval method called hybrid information descriptors (HIDs) consisting of mutual information descriptors (MIDs) and self information descriptors (SIDs). Based on the physiological structure of human eyes and visual perception mechanism, HIDs are designed to explore the internal correlations among different image feature spaces with image structure and multi-scale analysis, not only characterizing the low-level features, such as color, shape and texture, but also imitating the process of visual information transfer and perception in high-level understanding with the help of the proposed visual optimization model for feature fusion. Comparing with other existing methods applied to content-based image retrieval (CBIR) on four datasets, the usefulness and effectiveness of the HIDs are shown. Extensive experimental results can also demonstrate this.

A Content Based Image Retrieval Model Using Feature Space Dividing

This paper introduces a model for content based image retrieval. The proposed model extracts image color, texture and shape as feature vectors; and then the image feature space is divided into a group of search zones; during the image searching phase, the fractional order distance is utilized to evaluate the similarity between images. As the query image vector only needs to compare with library image vectors located in the same search zone, the time cost is largely reduced. Further more the fractional order distance is utilized to improve the vector matching accuracy. The experimental results demonstrated that the proposed model provides more accurate retrieval results with less time cost compared with other methods.

Leveraging local neighborhood topology for large scale person re-identification

In this paper we describe a semi-supervised approach to person re-identification that combines discriminative models of person identity with a Conditional Random Field (CRF) to exploit the local manifold approximation induced by the nearest neighbor graph in feature space. The linear discriminative models learned on few gallery images provides coarse separation of probe images into identities, while a graph topology defined by distances between all person images in feature space leverages local support for label propagation in the CRF. We evaluate our approach using multiple scenarios on several publicly available datasets, where the number of identities varies from 28 to 191 and the number of images ranges between 1003 and 36171. We demonstrate that the discriminative model and the CRF are complementary and that the combination of both leads to significant improvement over state-of-the-art approaches. We further demonstrate how the performance of our approach improves with increasing test data and also with increasing amounts of additional unlabeled data.

Multi-focus Image Fusion with Sparse Feature Based Pulse Coupled Neural Network

In order to better extract the focused regions and effectively improve the quality of the fused image, a novel multi-focus image fusion scheme with sparse feature based p ulse c oupled n eural n etwork ( PCNN ) is proposed. The registered source images are decomposed into principal matrices and sparse matrices by r obust p rincipal c omponent a nalysis ( RPCA ). T he salient features of the sparse matrices construct the sparse feature space of the source images. The sparse features are used to motivate the PCNN neurons. The focused regions of the source images are detected by the output of the PCNN and integrat ed to construct the final fused image. Experimental results show that the proposed scheme works better in extracting the focused regions and improving the fusion quality compared to the other existing fusion methods in both spatial and transform domain.

Fast Closed-Form Matting Using a Hierarchical Data Structure

Image/video matting is one of the key operations in many image/video editing applications. Although previous methods can generate high-quality matting results, their high computational cost in processing high-resolution image and video data often limits their usability. In this paper, we present a unified acceleration method for closed-form image and video matting using a hierarchical data structure, which achieves an excellent compromise between quality and speed. We first apply a Gaussian KD tree to adaptively cluster the input high-dimensional image and video feature space into a low-dimensional feature space. Then, we solve the affinity-weighted Laplacian alpha matting in the reduced feature space. The final matting results are derived using detail-aware alpha interpolation. Our algorithm can be fully parallelized by exploiting advanced graphics hardware, which can further accelerate the matting computation. Our method accelerates existing methods by at least an order of magnitude with good quality, and also greatly reduces the memory consumption. This acceleration strategy is also extended to support other affinity-based matting approaches, which makes it a more general accelerating framework for a variety of matting methods. Finally, we apply the presented method to accelerate image and video dehazing, and image shadow detection and removal.

Sparse Representation based Target Detection in Infrared Image

Infrared target detection research has recently attracted much attention, especially in pedestrian detection. In this paper, we use compressive sensing theory, a new and emerging field, for an infrared target detection system. Based on the framework of Bayesian filter, target is expressed by sparse representation. Compressive sensing (CS) is based on the illusion that a small quantity of non-adaptive linear projection of a compressible signal contains sufficient information for signal reconstruction and processing. In this paper we give an overview of compressive sensing and our proposed method. The method firstly construct appearance model using features extracted from the data independent image feature space. The appearance model can preserve structure targets’ feature since it adopts non-adaptive random projections. The experiment results show that the proposed infrared target detection system is feasible and efficient.