Lei Zhu

Human Tracking and Counting Using the KINECT Range Sensor Based on Adaboost and Kalman Filter

Conventional methods for human tracking and counting are based on images captured by 2-D frontal cameras, which have a major problem of occlusion among the people to be counted. In our paper, we use a 3-D sensor (Kinect) to capture the top-down view of the flow of people at the entrance of a premise for human counting purposes. In particular we use the Head and Shoulder Profile (HASP) of a human as the input feature. Then we use an Adaboost algorithm built from weak classifiers sensitive to certain spatial input features for detecting human objects from the input. Therefore, our system can detect a human facing all directions correctly. After detection, a Kalman based tracker is used to track the detected human object and filter false detection, which improves the false positive detection rate significantly. Our experiment result shows that the system can detect and track human motion accurately in real time at about 20 Frames per second.

Coarse-to-Fine Normal Filtering for Feature-Preserving Mesh Denoising Based on Isotropic Subneighborhoods

State-of-the-art normal filters usually denoise each face normal using its entire anisotropic neighborhood. However, enforcing these filters indiscriminately on the anisotropic neighborhood will lead to feature blurring, especially in challenging regions with shallow features. We develop a novel mesh denoising framework which can effectively preserve features with various sizes. Our idea is inspired by the observation that the underlying surface of a noisy mesh is piecewise smooth. In this regard, it is more desirable that we denoise each face normal within its piecewise smooth region (we call such a region as an isotropic subneighborhood) instead of using the anisotropic neighborhood. To achieve this, we first classify mesh faces into several types using a face normal tensor voting and then perform a normal filter to obtain a denoised coarse normal field. Based on the results of normal classification and the denoised coarse normal field, we segment the anisotropic neighborhood of every feature face into a number of isotropic subneighborhoods via local spectral clustering. Thus face normal filtering can be performed again on the isotropic subneighborhoods and produce a more accurate normal field. Extensive tests on various models demonstrate that our method can achieve better performance than state-of-the-art normal filters, especially in challenging regions with features.

Detection and measurement of fetal abdominal contour in ultrasound images via local phase information and iterative randomized Hough transform.

Due to the characteristic artifacts of ultrasound images, e.g., speckle noise, shadows and intensity inhomogeneity, traditional intensity-based methods usually have limited success on the segmentation of fetal abdominal contour. This paper presents a novel approach to detect and measure the abdominal contour from fetal ultrasound images in two steps. First, a local phase-based measure called multiscale feature asymmetry (MSFA) is de ned from the monogenic signal to detect the boundaries of fetal abdomen. The MSFA measure is intensity invariant and provides an absolute measurement for the signi cance of features in the image. Second, in order to detect the ellipse that ts to the abdominal contour, the iterative randomized Hough transform is employed to exclude the interferences of the inner boundaries, after which the detected ellipse gradually converges to the outer boundaries of the abdomen. Experimental results in clinical ultrasound images demonstrate the high agreement between our approach and manual approach on the measurement of abdominal circumference (mean sign difference is 0.42% and correlation coef cient is 0.9973), which indicates that the proposed approach can be used as a reliable and accurate tool for obstetrical care and diagnosis.

Fast feature-preserving speckle reduction for ultrasound images via phase congruency

Ultrasonography is widely used in clinical diagnosis and therapeutic procedures, but speckle noise often obscures important features and complicates interpretation and analysis of ultrasound images. In this regard, speckle reduction is a crucial prerequisite of many computer aided ultrasound diagnosis and treatment systems However, removing speckle noise while simultaneously preserving features in ultrasound images is a challenging task. We propose a novel optimization framework for speckle reduction by leveraging the concept of phase congruency and incorporating a feature asymmetry metric into the regularization term of the objective function to effectively distinguish the features and speckle noise. The feature asymmetry metric can productively separate features from speckle noise by analyzing the local frequency information. Compared with traditional methods employing intensity gradients as regularization terms, our framework is invariant to the intensity amplitude of features so that low contrast features are almost equally protected as high contrast features. In addition, rather than adopting the gradient descent, we propose a novel solver by decomposing the original non-convex optimization into solving several linear systems, leading to an efficient solution of the optimization. Owing to different penalties on speckle noise and features, our method can efficiently remove speckle noise and preserve features at the same time. Experiments on simulated and real ultrasound images demonstrate our method can better maintain features with speckle removal than state-of-the-art methods, especially for the low contrast features. HighlightsWe propose a novel optimization framework for speckle reduction by leveraging the concept of phase congruency and incorporating a feature asymmetry metric into the regularization term of the objective function to effectively distinguish the features and speckle noise.We propose a novel solver by decomposing the original nonconvex optimization into solving several linear systems, leading to an efficient solution of the optimization.Compared with traditional methods employing intensity gradients as regularization terms, our framework is invariant to the intensity amplitude of features so that low contrast features are almost equally protected as high contrast features.

A Non-local Low-Rank Framework for Ultrasound Speckle Reduction

Speckle refers to the granular patterns that occur in ultrasound images due to wave interference. Speckle removal can greatly improve the visibility of the underlying structures in an ultrasound image and enhance subsequent post processing. We present a novel framework for speckle removal based on low-rank non-local filtering. Our approach works by first computing a guidance image that assists in the selection of candidate patches for non-local filtering in the face of significant speckles. The candidate patches are further refined using a low-rank minimization estimated using a truncated weighted nuclear norm (TWNN) and structured sparsity. We show that the proposed filtering framework produces results that outperform state-of-the-art methods both qualitatively and quantitatively. This framework also provides better segmentation results when used for pre-processing ultrasound images.

Non-local sparse and low-rank regularization for structure-preserving image smoothing

This paper presents a new image smoothing method that better preserves prominent structures. Our method is inspired by the recent non‐local image processing techniques on the patch grouping and filtering. Overall, it has three major contributions over previous works. First, we employ the diffusion map as the guidance image to improve the accuracy of patch similarity estimation using the region covariance descriptor. Second, we model structure‐preserving image smoothing as a low‐rank matrix recovery problem, aiming at effectively filtering the texture information in similar patches. Lastly, we devise an objective function, namely the weighted robust principle component analysis (WRPCA), by regularizing the low rank with the weighted nuclear norm and sparsity pursuit with L1 norm, and solve this non‐convex WRPCA optimization problem by adopting the alternative direction method of multipliers (ADMM) technique. We experiment our method with a wide variety of images and compare it against several state‐of‐the‐art methods. The results show that our method achieves better structure preservation and texture suppression as compared to other methods. We also show the applicability of our method on several image processing tasks such as edge detection, texture enhancement and seam carving.

Effective mesh smoothing for haptic rendering in medical applications

Mesh models can be extracted from medical imaging data. However some methods (e.g., CT) may suffer from severe artifacts (e.g., staircases, noises) in current clinical routine. As a consequence, haptic systems, when using these influenced mesh models, will become unstable. To tackle this problem, in this paper we propose an effective medical-oriented smoothing algorithm focusing on haptic rendering. Our algorithm mainly consists of two stages, namely vertex re-sampling and surface fitting. The first stage is adopted to eliminate staircases while the second can obtain the underlying surface by least square fitting method. Experiments on various medical imaging data present the efficacy of our methodology, which can achieve higher quality results than previous approaches regarding both surface smoothness and surface accuracy. And the final results on haptic applications further show this proposed technique is suitable for medical surgery simulations.

Phase-based feature detection in fetal ultrasound images

Detection of image features is an essential step in many medical applications. However, it is very challenging to accurately extract important features from ultrasound data that is corrupted by various imaging artifacts. Traditional intensity-based methods generally have poor performance in detecting salient features from ultrasound images. In contrast, phase-base approaches have been shown to perform well in these images because they are theoretically intensity invariant. In this paper, we extend previous phase-based methods to the field of fetal ultrasound images to detect both symmetric and asymmetric features, which correspond to ridge-like and step edge-like object boundaries, respectively. This is achieved by exploiting local phase-based measures computed from a 2D isotropic analytic signal: monogenic signal. Experimental results in clinical images demonstrate the outperformance of the proposed approach.

Deep Attentional Features for Prostate Segmentation in Ultrasound

Automatic prostate segmentation in transrectal ultrasound (TRUS) is of essential importance for image-guided prostate biopsy and treatment planning. However, developing such automatic solutions remains very challenging due to the ambiguous boundary and inhomogeneous intensity distribution of the prostate in TRUS. This paper develops a novel deep neural network equipped with deep attentional feature (DAF) modules for better prostate segmentation in TRUS by fully exploiting the complementary information encoded in different layers of the convolutional neural network (CNN). Our DAF utilizes the attention mechanism to selectively leverage the multi-level features integrated from different layers to refine the features at each individual layer, suppressing the non-prostate noise at shallow layers of the CNN and increasing more prostate details into features at deep layers. We evaluate the efficacy of the proposed network on challenging prostate TRUS images, and the experimental results demonstrate that our network outperforms state-of-the-art methods by a large margin.

R^3Net: Recurrent Residual Refinement Network for Saliency Detection

Saliency detection is a fundamental yet challenging task in computer vision, aiming at highlighting the most visually distinctive objects in an image. We propose a novel recurrent residual refinement network (RNet) equipped with residual refinement blocks (RRBs) to more accurately detect salient regions of an input image. Our RRBs learn the residual between the intermediate saliency prediction and the ground truth by alternatively leveraging the low-level integrated features and the highlevel integrated features of a fully convolutional network (FCN). While the low-level integrated features are capable of capturing more saliency details, the high-level integrated features can reduce non-salient regions in the intermediate prediction. Furthermore, the RRBs can obtain complementary saliency information of the intermediate prediction, and add the residual into the intermediate prediction to refine the saliency maps. We evaluate the proposed RNet on five widely-used saliency detection benchmarks by comparing it with 16 stateof-the-art saliency detectors. Experimental results show that our network outperforms our competitors in all the benchmark datasets.