Zhiguo Cao

Fast new small-target detection algorithm based on a modified partial differential equation in infrared clutter

To detect and track moving dim targets against the complex cluttered background in IR image sequences is still a difficult problem because the nonstationary structured background clutter usually results in low target detectability and a high probability of false alarm. A new adaptive anisotropic filter based on a modified partial differential equation (AFMPDE) is proposed to detect a small target in such a strong cluttered background. A regularizing operator is employed to adaptively eliminate structured background and simultaneously enhance the target signal. The proposed algorithms performance is illustrated and compared with the two-dimensional least mean square (TDLMS) adaptive filter on real IR image data. Experimental results demonstrate that the proposed novel method is fast and effective.

A real-time embedded architecture for SIFT

SIFT has shown a great success in various computer vision applications. However, its large computational complexity has been a challenge to most embedded implementations. This paper presents a low-cost embedded system based on a new architecture that successfully integrates FPGA and DSP. It optimizes the FPGA architecture for the feature detection step of SIFT to reduce the resource utilization, and optimizes the implementation of the feature description step using a high-performance DSP. Due to this novel design, this system can detect SIFT feature and extract SIFT descriptor for detected features in real-time. Extensive experiments demonstrate its effectiveness and efficiency.

An Embedded System-on-Chip Architecture for Real-time Visual Detection and Matching

Detecting and matching image features is a fundamental task in video analytics and computer vision systems. It establishes the correspondences between two images taken at different time instants or from different viewpoints. However, its large computational complexity has been a challenge to most embedded systems. This paper proposes a new FPGA-based embedded system architecture for feature detection and matching. It consists of scale-invariant feature transform (SIFT) feature detection, as well as binary robust independent elementary features (BRIEF) feature description and matching. It is able to establish accurate correspondences between consecutive frames for 720-p (1280x720) video. It optimizes the FPGA architecture for the SIFT feature detection to reduce the utilization of FPGA resources. Moreover, it implements the BRIEF feature description and matching on FPGA. Due to these contributions, the proposed system achieves feature detection and matching at 60 frame/s for 720-p video. Its processing speed can meet and even exceed the demand of most real-life real-time video analytics applications. Extensive experiments have demonstrated its efficiency and effectiveness.

A multisize superpixel approach for salient object detection based on multivariate normal distribution estimation.

This paper presents a new method for salient object detection based on a sophisticated appearance comparison of multisize superpixels. Those superpixels are modeled by multivariate normal distributions in CIE-Lab color space, which are estimated from the pixels they comprise. This fitting facilitates an efficient application of the Wasserstein distance on the Euclidean norm (W2) to measure perceptual similarity between elements. Saliency is computed in two ways. On the one hand, we compute global saliency by probabilistically grouping visually similar superpixels into clusters and rate their compactness. On the other hand, we use the same distance measure to determine local center-surround contrasts between superpixels. Then, an innovative locally constrained random walk technique that considers local similarity between elements balances the saliency ratings inside probable objects and background. The results of our experiments show the robustness and efficiency of our approach against 11 recently published state-of-the-art saliency detection methods on five widely used benchmark data sets.

A fast and robust local descriptor for 3D point cloud registration

This paper proposes a novel local feature descriptor, called a local feature statistics histogram (LFSH), for efficient 3D point cloud registration. An LFSH forms a comprehensive description of local shape geometries by encoding their statistical properties on local depth, point density, and angles between normals. The sub-features in the LFSH descriptor are low-dimensional and quite efficient to compute. In addition, an optimized sample consensus (OSAC) algorithm is developed to iteratively estimate the optimum transformation from point correspondences. OSAC can handle the challenging cases of matching highly self-similar models. Based on the proposed LFSH and OSAC, a coarse-to-fine algorithm can be formed for 3D point cloud registration. Experiments and comparisons with the state-of-the-art descriptors demonstrate that LFSH is highly discriminative, robust, and significantly faster than other descriptors. Meanwhile, the proposed coarse-to-fine registration algorithm is demonstrated to be robust to common nuisances, including noise and varying point cloud resolutions, and can achieve high accuracy on both model data and scene data.

Cloud Classification of Ground-Based Images Using Texture–Structure Features

AbstractCloud classification of ground-based images is a challenging task. Recent research has focused on extracting discriminative image features, which are mainly divided into two categories: 1) choosing appropriate texture features and 2) constructing structure features. However, simply using texture or structure features separately may not produce a high performance for cloud classification. In this paper, an algorithm is proposed that can capture both texture and structure information from a color sky image. The algorithm comprises three main stages. First, a preprocessing color census transform (CCT) is applied. The CCT contains two steps: converting red, green, and blue (RGB) values to opponent color space and applying census transform to each component. The CCT can capture texture and local structure information. Second, a novel automatic block assignment method is proposed that can capture global rough structure information. A histogram and image statistics are computed in every block and are con...

Entropic thresholding based on gray-level spatial correlation histogram

In this paper, an entropic thresholding method based on the gray-level spatial correlation (GLSC) histogram defined by ourselves is presented. Compared with traditional two-dimensional histogram, we take into account the image local property in a different way by GLSC histogram. In experiment, we make comparison of the proposed method with two-dimensional entropic thresholding method proposed by Abutaleb and one-dimensional entropic thresholding method proposed by Kapur. The experiment demonstrates that generally our method could yield equivalent or even better result than Abutalebpsilas method while saving time remarkably and perform much better than Kapurpsilas method without too more time consumption.

Type-2 fuzzy thresholding using GLSC histogram of human visual nonlinearity characteristics

Image thresholding is one of the most important approaches for image segmentation and it has been extensively used in many image processing or computer vision applications. In this paper, a new image thresholding method is presented using type-2 fuzzy sets based on GLSC histogram of human visual nonlinearity characteristics (HVNC).The traditional GLSC histogram takes the image spatial information into account in a different way from two-dimensional histogram. This work refines the GLSC histogram by embedding HVNC into GLSC histogram. To select threshold based on the redefined GLSC histogram, we employ the type-2 fuzzy set, whose membership function integrates the effect of pixel gray value and local spatial information to membership value. The type-2 fuzzy set is subsequently transformed into a type-1 fuzzy set for fuzziness measure computation via type reduction. Finally, the optimal threshold is obtained by minimizing the fuzziness of the type-1 fuzzy set after an exhaustive search. The experiment on different types of images demonstrates the effectiveness and the robustness of our proposed thresholding technique.

Entropic image thresholding based on GLGM histogram

We propose GLGM (gray-level & gradient-magnitude) histogram as a novel image histogram for thresholding. GLGM histogram explicitly captures the gray level occurrence probability and spatial distribution property simultaneously. Different from previous histograms that also consider the spatial information, GLGM histogram employs the Fibonacci quantized gradient magnitude to characterize spatial information effectively. In this paper, it is applied to entropic image thresholding. For threshold selection, we define a new spatial property weighting function to depict the roles played by different kinds of pixels. The experiments demonstrate the effectiveness and robustness of our thresholding approach, containing wide range comparisons with the well established thresholding methods.

An improved MRF-based change detection approach for multitemporal remote sensing imagery

In the task of multitemporal remote sensing image change detection, conventional Markov random field (MRF) based approaches consider contextual information between neighboring pixels to obtain the change map. However, these approaches often get erroneous results at discontinuities such as edges, ridges and valleys, since they assume that neighboring pixels tend to have the same label. To overcome this, an improved MRF based change detection approach for multitemporal remote sensing imagery is proposed. The method first finds edges in the difference image by using the line process. Then, the weights of MRF prior energy are adaptively adjusted by considering the gray level differences between neighboring pixels. A group of adaptive weighting functions are defined in the study, and their performances in the task of change detection are compared. Experimental results confirm the proposed approach.