Yuzhong Shen

Fuzzy vector median-based surface smoothing

We propose a novel approach for smoothing surfaces represented by triangular meshes. The proposed method is a two-step procedure: surface normal smoothing through fuzzy vector median (FVM) filtering followed by integration of surface normals for vertex position update based on the least square error (LSE) criteria. Median and order statistic-based filters are extensively used in signal processing, especially image processing, due to their ability to reject outliers and preserve features such as edges and monotonic regions. More recently, fuzzy ordering theory has been introduced to allow averaging among similarly valued samples. Fuzzy ordering theory leads naturally to the fuzzy median, which yields improved noise smoothing over traditional crisp median filters. We extend the fuzzy ordering concept to vector-based data and introduce the fuzzy vector median filter. The application of FVM filters to surface normal smoothing yields improved results over previously introduced normal smoothing algorithms. The improved filtering results, coupled with LSE vertex position update, produces surface smoothing that minimizes the effects of noise while simultaneously preserving detail features. The proposed method is simple to implement and relatively fast. Simulation results are presented showing the performance of the proposed method and its advantages over commonly used surface smoothing algorithms. Additionally, optimization procedures for FVM filters are derived and evaluated.

Noise reduction and edge detection via kernel anisotropic diffusion

A novel kernel anisotropic diffusion (KAD) method is proposed for robust noise reduction and edge detection. The KAD incorporates a kernelized gradient operator in the diffusion, leading to more effective edge detection and providing a better control to the diffusion process. Adaptive diffusion threshold estimation and automatic diffusion termination criterion are also introduced to enhance the robustness of the KAD. The KAD outperforms several previous anisotropic diffusion-based methods for low SNR images.

Fast adaptive optimization of weighted vector median filters

Weighted vector median (WVM) filters are effective tools for multichannel signal processing. To obtain the desired filtering behavior and characteristic, the WVM filter weights must be determined in an appropriate manner. In this paper, we first analyze previously defined approaches for WVM filter optimization and show their drawbacks related to derivative computation and vector direction information utilization. Based on this analysis, we propose two fast adaptive algorithms for WVM filter design. Proposed Algorithm I computes locally optimal weight changes at each iteration and updates the filter weights accordingly. This algorithm does not involve derivative computation, thus eliminating the instability caused by derivative approximations utilized in previous approaches. Proposed Algorithm II extends the results from established marginal weighted median optimization methods to the vector case by error metric generalization. Both algorithms can be applied to WVM filters using the Lp norm, while Algorithm I can operate on more general distance metrics. The presented simulation results show that both algorithms are effective, fast, and stable; they perform well under a wide range of circumstances

Speckle reduction of ultrasound images based on Rayleigh-trimmed anisotropic diffusion filter

A novel method is proposed to reduce speckle in ultrasound images. Based on the assumption of Rayleigh distribution of speckle, a Rayleigh-trimmed filter is first proposed to estimate the relative standard deviations of local signals and the results are used to determine the parameter that controls an alpha-trimmed mean filter for suppressing the primary noise. Then the anisotropic diffusion is subsequently applied to further reduce noise while enhancing features and structures in the original image. We also extend the proposed method to three-dimensional space by introducing time as one additional dimension. The proposed method effectively utilizes the statistical characteristics of speckle and the two-step despeckling algorithm reduces speckle significantly while retaining important features. The effectiveness of the proposed method is well demonstrated by experiments on both simulated and real ultrasound images.

Energy Evaluation for Applications with Different Thread Affinities on the Intel Xeon Phi

The Intel Xeon Phi coprocessor offers high parallelism on energy-efficient hardware to minimize energy consumption while maintaining performance. Dynamic frequency and voltage scaling is not accessible on the Intel Xeon Phi. Hence, saving energy relies mainly on tuning application performance. One general optimization technique is thread affinity, which is an important factor in multi-core architectures. This work investigates the effects of varying thread affinity modes and reducing core utilization on energy and execution time for the NASA Advanced Supercomputing Parallel Benchmarks (NPB). Energy measurements are captured using the micsmc utility tool available on Xeon Phi. The measurements are checked against total power captured using Watts up power meters. The results are compared to the system-default thread affinity and granularity modes. Mostly positive impacts on performance and energy are observed: When executed at the maximum thread count on all unoccupied cores, all the benchmarks but one exhibited energy savings if a specific affinity mode is set.

Modeling performance and energy for applications offloaded to Intel Xeon Phi

Accelerators are adopted to increase performance, reduce time-to-solution, and minimize energy-to-solution. However, employing them efficiently, given system and application characteristics, is often a daunting task. A goal of this work is to propose a general model that predicts performance and power requirements for an application, computational portions of which are offloaded to an accelerator. Intel Xeon Phi is the only accelerator type investigated here, and only in offload execution mode. This mode is also employed by other accelerator types, such as GPU; thus the proposed model is applicable directly. The predictive capabilities of the model are demonstrated by determining the best hardware-software configuration instances with respect to the minimum energy consumption for the CoMD proxy application executed on single or multiple nodes. For the CoMD problem sizes investigated here, the best modeled configuration was relatively close to the best measured configuration with relative error under 5% of the energy consumed for most configurations. Initial model validation also confirmed the model accuracy for a variety of model parameters, such as host computation time and power consumption on the host and accelerator. The model also provides estimates of the total data movement and computational throughput as well as of some key metrics, such as FLOPs-per-joule and bytes-per-joule, which are commonly used to study the energy-performance trade-offs.

Automatic high-fidelity 3D road network modeling based on 2D GIS data

Abstract Many computer applications such as racing games and driving simulations demand high-fidelity 3D road network models. However, few methods exist for the automatic generation of 3D realistic road networks, especially for those in the real world. On the other hand, vast 2D road network data in various geographical information systems (GIS) have been collected in the past and are used by a wide range of applications. A method that can automatically produce 3D high-fidelity road network models from 2D real road GIS data will significantly reduce both the labor and time cost, and greatly benefit applications involving road networks. Based on a set of carefully selected civil engineering rules for road design, this paper proposes a novel approach that transforms existing road GIS data that contain only 2D road centerline information into high-fidelity 3D road network models. The proposed method consists of several major components, including road GIS data preprocessing, 3D centerline modeling, and 3D geometric modeling. With this approach, basic road elements such as road segments, road intersections and traffic interchanges are generated automatically to compose sophisticated road networks in a seamless manner. Results show that this approach provides a rapid and efficient 3D road modeling method for applications that have stringent requirements on high-fidelity road models.

An automated diagnostic system of polycystic ovary syndrome based on object growing

OBJECTIVE Polycystic ovary syndrome (PCOS) is a complex endocrine disorder that seriously affects womens health. The disorder is characterized by the formation of many follicles in the ovary. Currently the predominant diagnosis is to manually count the number of follicles, which may lead to inter-observer and intra-observer variability and low efficiency. A computer-aided PCOS diagnostic system may overcome these problems. However the methods reported in recently published literature are not very effective and often need human interaction. To overcome these problems, we propose an automated PCOS diagnostic system based on ultrasound images. METHODS AND MATERIALS The proposed system consists of two major functional blocks: preprocessing phase and follicle identification based on object growing. In the preprocessing phase, speckle noise in the input image is removed by an adaptive morphological filter, then contours of objects are extracted using an enhanced labeled watershed algorithm, and finally the region of interest is automatically selected. The object growing algorithm for follicle identification first computes a cost map to distinguish between the ovary and its external region and assigns each object a cost function based on the cost map. The object growing algorithm initially selects several objects that are likely to be follicles with very high probabilities and dynamically update the set of possible follicles based on their cost functions. The proposed method was applied to 31 real PCOS ultrasound images obtained from patients and its performance was compared with those of three other methods, including level set method, boundary vector field (BVF) method and the fuzzy support vector machine (FSVM) classifier. RESULTS Based on the judgment of subject matter experts, the proposed diagnostic system achieved 89.4% recognition rate (RR) and 7.45% misidentification rate (MR) while the RR and MR of the level set method, the BVF method and the FSVM classifier are around 65.3% and 2.11%, 76.1% and 4.53%, and 84.0% and 16.3%, respectively. The proposed diagnostic system also achieved better performance than those reported in recently published literature. CONCLUSION The paper proposed an automated diagnostic system for the PCOS using ultrasound images, which consists of two major functional blocks: preprocessing phase and follicle identification based on object growing. Experimental results showed that the proposed system is very effective in follicle identification for PCOS diagnosis.

Runtime power limiting of parallel applications on Intel Xeon Phi processors

Energy-efficient computing is crucial to achieving exascale performance. Power capping and dynamic voltage/frequency scaling may be used to achieve energy savings. The Intel Xeon Phi implements a power capping strategy, where power thresholds are employed to dynamically set voltage/frequency at the runtime. By default, these power limits are much higher than the majority of applications would reach. Hence, this work aims to set the power limits according to the workload characteristics and application performance. Certain models, originally developed for the CPU performance and power, have been adapted here to determine power-limit thresholds in the Xeon Phi. Next, a procedure to select these thresholds dynamically is proposed, and its limitations outlined. When this runtime procedure along with static power-threshold assignment were compared with the default execution, energy savings ranging from 5% to 49% were observed, mostly for memory-intensive applications.

Active cardiac model and its application on structure detection from early fetal ultrasound sequences

The structure of an early fetal heart provides vital information for the diagnosis of fetus defects. However, early fetal hearts are difficult to detect due to their relatively small size and the low signal-to-noise ratio of ultrasound images. In this paper, a novel method is proposed for automatic detection of early fetal cardiac structure from ultrasound images. The proposed method consists of two major parts which are the preprocessing phase and the active cardiac model: (1) The preprocessing phase consists of two sub-steps. (a) The region of interest is first automatically selected based on an accumulated motion image, which is able to represent the motion information of the fetal heart more accurately. (b) Then by combining Rayleigh-trimmed filter and anisotropic diffusion in 3-dimensional space, a despeckling method is developed to suppress the speckle noise and emphasize the motion information for subsequent cardiac structure detection. (2) The active cardiac model is proposed for the detection of fetal heart structure, which is a key contribution of this paper. It takes into account both the structure and motion information of fetal hearts simultaneously. Both learning and inference of the active cardiac model are described in the paper. Experiments on seven ultrasound sequences demonstrate the effectiveness of the proposed method.