Human linear template with mammographic backgrounds estimated with a genetic algorithm

We estimated human observer linear templates underlying the detection of a realistic, spherical mass signal with mammographic backgrounds. Five trained na?ve observers participated in two-alternative forced-choice (2-AFC) detection experiments with the signal superimposed on synthetic, clustered lumpy backgrounds (CLBs) in one condition and on nonstationary real mammographic backgrounds in another. Human observer linear templates were estimated using a genetic algorithm. A variety of common model observer templates were computed, and their shapes and associated performances were compared with those of the human observer. The estimated linear templates are not significantly different for stationary CLBs and real mammographic backgrounds. The estimated performance of the linear template compared with that of the human observers is within 5% in terms of percent correct (Pc) for the 2-AFC task. Channelized Hotelling models can fit human performance, but the templates differ considerably from the human linear template. Due to different local statistics, detection efficiency is significantly higher on nonstationary real backgrounds than on globally stationary synthetic CLBs. This finding emphasizes that nonstationary backgrounds need to be described by their local statistics.     

Visual signal detection in structured backgrounds. III. Calculation of figures of merit for model observers in statistically nonstationary backgrounds

Models of human visual detection have been successfully used in computer-generated noise. For these backgrounds, which are generally statistically stationary, model performance can be readily calculated by computing the index of detectability d' from the noise power spectrum, the signal profile, and the model template. However, model observers are ultimately needed in more real backgrounds, which may be statistically non-stationary. We investigated different methods to calculate figures of merit for model observers in real backgrounds based on different assumptions about image stationarity. We computed performance of the nonpre-whitening matched-filter observer with an eye filter on mammography and coronary angiography for an additive or a multiplicative signal. Performance was measured either by applying the model template to the images or by computing closed-form expressions with various assumptions about image stationarity. Results show first that the structured backgrounds investigated cannot be considered stationary. Second, traditional closed-form expressions of detectability calculated from the noise power spectra with the assumption of background stationarity lead to erroneous estimates of model performance. Third, the most accurate way of measuring model performances is by directly applying the model template on the images or by computing a closed-form expression that does not assume image stationarity.     

Visual signal detection with two-component noise: low-pass spectrum effects

Detection of signals in natural images and scenes is limited by both noise and structure. The purpose of this study is to investigate phenomenological issues of signal detection in two-component noise. One component had a broadband (white) spectrum designed to simulate image noise. The other component was filtered to simulate two classes of low-pass background structure spectra: Gaussian-filtered noise and power-law noise. Measurements of human and model observer performance are reported for several aperiodic signals and both classes of background spectra. Human results are compared with two classes of observer models and are fitted very well by suboptimal prewhitening matched filter models. The nonprewhitening model with an eye filter does not agree with human results when background-noise-component power spectrum bandwidths are less than signal energy bandwidths.     

Is motion processing unitary?

Motion discrimination space is conventionally categorized into motion detection, speed discrimination, and direction discrimination tasks. But an ideal observer uses a unitary motion mechanism that is affected only by the noise level and the difference in speed (or displacement) between two stimuli. We tested whether human performance in the various motion tasks showed the working of a unitary mechanism or the combined outputs of more than one mechanism. We examined the whole motion discrimination space, using random dots that underwent a sudden jump or displacement. The discriminability was measured as a function of the standard and comparison displacements. Both the ideal observer model and a nonideal observer model that contains additive internal noise predict a planar response surface. When the dot motion was noiseless, the planar surface fitted well except for much higher than expected sensitivity for motion detection. This is consistent with a purely temporal mechanism that uses flicker or a purely spatial mechanism that uses the length of time-averaged streaks. It is also consistent with a Weber's law device. When motion noise was added to the displays, the planar response surface again fitted well, although the residuals showed the presence of a speed energy mechanism. We conclude that a unitary motion mechanism exists (nonideal observer model), although its performance may be supplemented by other mechanisms whose main impact is on discrimination of speeds near zero.     

Temporal detection in human vision: dependence on spatial frequency.

In the study of perception of temporal changes in luminance, it is customary to model perceptual performance as based on one or more linear filters. The task is then to estimate the temporal impulse responses or the representation of the impulse response in the frequency domain. Previously, temporal masking data have been used to estimate the properties and numbers of these temporal mechanisms (filters) in central vision for 1-cycle-per-degree (cpd) targets [Vision Res. 38, 1023 (1998)]. The same methods have been used to explore how properties of the estimated filters change with stimulus contrast energy [J. Opt. Soc. Am. A 14, 2557 (1997)]. We present estimated properties for temporal mechanisms that detect low spatial-frequency patterns. The results indicate that two filters provide the best model for performance when mask contrast is significant. There are also differences between properties for mechanisms that detect signal spatial frequencies of 1 cpd and 1/3 cpd. The sensitivity of the low-pass mechanism relative to the bandpass mechanism is reduced at 1/3 cpd, consistent with previous findings.     

Task-based lens design with application to digital mammography

Recent advances in model observers that predict human perceptual performance now make it possible to optimize medical imaging systems for human task performance. We illustrate the procedure by considering the design of a lens for use in an optically coupled digital mammography system. The channelized Hotelling observer is used to model human performance, and the channels chosen are differences of Gaussians. The task performed by the model observer is detection of a lesion at a random but known location in a clustered lumpy background mimicking breast tissue. The entire system is simulated with a Monte Carlo application according to physics principles, and the main system component under study is the imaging lens that couples a fluorescent screen to a CCD detector. The signal-to-noise ratio (SNR) of the channelized Hotelling observer is used to quantify this detectability of the simulated lesion (signal) on the simulated mammographic background. Plots of channelized Hotelling SNR versus signal location for various lens apertures, various working distances, and various focusing places are presented. These plots thus illustrate the trade-off between coupling efficiency and blur in a task-based manner. In this way, the channelized Hotelling SNR is used as a merit function for lens design.     

改进型干扰观测器在稳定平台伺服系统中的应用

为提高稳定平台伺服系统对各种扰动的抑制性能,将一种改进型干扰观测器应用到了平台系统中,该改进型干扰观测器在经典干扰观测器的基础上新增了一个补偿系统输出的控制结构。文中对改进型干扰观测的性能进行了详细分析,并对存在模型摄动时的鲁棒稳定性进行了推导,同时给出了低通滤波器 的设计原则。最后通过数值仿真分析表明,改进型干扰观测器能够较好地消除外部干扰和模型摄动的影响,改善了测量噪声的抑制性能。     

基于噪声模板与互相关计算的涡街流量计抗强振动方法

根据涡街流量计在工作现场受强振动噪声干扰的情况,搭建振动实验平台,进行振动实验。对采集的数据进行功率谱分析和带宽特征分析,建立噪声模板,并量化描述噪声的带宽特性。进行噪声模板与传感器输出信号功率谱的互相关计算,提出利用噪声和涡街信号归一化互相关系数曲线进行噪声判别的抗强振动干扰的方法。对振动实验中采集的传感器数据进行处理,验证了该方法的有效性。     

Ideal-observer computation in medical imaging with use of Markov-chain Monte Carlo techniques

The ideal observer sets an upper limit on the performance of an observer on a detection or classification task. The performance of the ideal observer can be used to optimize hardware components of imaging systems and also to determine another observer's relative performance in comparison with the best possible observer. The ideal observer employs complete knowledge of the statistics of the imaging system, including the noise and object variability. Thus computing the ideal observer for images (large-dimensional vectors) is burdensome without severely restricting the randomness in the imaging system, e.g., assuming a flat object. We present a method for computing the ideal-observer test statistic and performance by using Markov-chain Monte Carlo techniques when we have a well-characterized imaging system, knowledge of the noise statistics, and a stochastic object model. We demonstrate the method by comparing three different parallel-hole collimator imaging systems in simulation.     

Detectability in nuclear medicine images by a random-dot model: a theoretical analysis

The signal-containing random-dot model is applied to an evaluation of theoretical detectabilities of computer-generated random-dot pattern images that are idealized nuclear-medicine images. Theoretical detectabilities, which can be obtained by a signal-containing random-dot model, are calculated for two simulated observation procedures of these patterns. One simulates the observation procedure in which the observer compares the transmittance at a center of the signal and the transmittance at a background-noise region far away from the signal. The other simulates the observation procedure in which the observer notices the transmittance difference across the boundary between the signal and the background noise. The theoretical detectability is compared with the result of the receiver operating characteristics analysis. The results of receiver operating characteristics analysis are in good agreement with the theoretical detectability obtained by the observation procedure in which the observer notices the transmittance difference across the boundary between the signal and the background noise.     

Effect of spatial frequency on the visibility of unstructured patterns.

Threshold measurements have been made on random patterns that have been computer filtered to contain only certain spatial frequencies. The dependence of threshold on the spatial-frequency distribution of the energy in the patterns was measured. The results indicate that the visibility of an unstructured pattern is dependent on the spatial-frequency content of the pattern. Near threshold, narrow-band noise is much more visible than wide-band noise. The narrow-band noise visibility is caused by the presence of structure in the pattern, some of which is rotation sensitive. For wider-band noise, the modulation transfer function is circularly symmetric. The visibility of very-wide-band noise near threshold depends only on the luminance variations concentrated within a one-octave spatial-frequency band.     

具有扰动观测器的汽车主动悬架滑模控制

基于双横臂悬架的1/4汽车模型,以降低簧载质量振动为目的,提出了具有扰动观测器的滑模控制方法。由1/4汽车模型中簧载质量的受力平衡条件,得到了1/4汽车模型的二阶线性控制系统。在路面激励位移未知时,为了估计二阶线性控制系统输入的扰动力(控制臂的作用力、悬架弹簧力、减振器阻尼力和外界干扰力的总和),给出了基于滑模控制的扰动观测器设计方法。根据扰动观测器的估计值,计算了具有扰动观测器的滑模控制的1/4汽车模型的控制力。当路面激励为阶跃位移时,计算分析了采用无控制、PID控制和滑模控制时1/4汽车模型的簧载质量加速度,结果表明:具有扰动观测器的滑模控制器可以获得更好的控制效果。通过分析不同控制参数和外界干扰力对滑模控制器效果的影响,证明了基于1/4汽车模型的具有扰动观测器的悬架滑模控制设计的有效性。     

An ideal observer with channels versus feature-independent processing of spatial frequency and orientation in visual search performance

An influential assumption for the front end of models in vision, visual search, and object recognition is an analysis of independent features that correspond to basic image properties, such as motion, shape, and color. Empirically, one common test of independent features (a cue-summation study) measures performance with increasing available cues or features, with improving performance leading to conclusions of summation across independent features. In a study by Shimozaki et al. [J. Vision 2, 354-370 (2002)], both ideal and human observers showed no summation with large stimulus differences, in contrast to independent-feature models and suggesting that stimulus information (as assessed by an ideal observer) might affect cue-summation studies. Extending the previous summation study, observers performed a visual search of four Gabors differing in only orientation, only spatial frequency, or both orientation and spatial frequency, across a range of target-distractor differences. An ideal observer underpredicted human summation for small differences, whereas the independent-orientation and spatial-frequency feature models overpredicted human summation for large differences. An ideal observer with channels jointly tuned to spatial frequency and orientation predicted human performance across both small and large target-distractor differences.     

Analysis of optical near-field images by Karhunen-Loève transformation

With regard to classical image transformations, the Karhunen-Loève transform is based on the statistical analysis of signal or image data. We propose using it to analyze and recognize the content of images generated in near-field microscopy. It is shown that such a transform is an efficient tool for separating actual information from noise without reducing the spatial-frequency band.     

深度神经网络在螺旋桨叶片数识别中的应用

从调制(Demodulation on Noise, DEMON)谱谐波簇中提取的结构特征可以建立用于螺旋桨叶片数识别的模板。使用模板匹配算法进行螺旋桨叶片数识别时,存在依赖模板库和置信度准则、算法约束条件多、无法发现缺失模板等问题。本文提出了一种将深度神经网络(Deep Neural Network, DNN)应用于螺旋桨叶片数识别的方法,该方法仅在训练深度神经网络时使用模板库,克服了识别过程中对模板库和置信度准则的依赖。此外,通过提取识别错误项,可以找到缺失模板,实现了对模板库数据的补充。使用该算法对大量实测数据进行检测,发现深度神经网络具有更高的识别正确率,而且识别过程更加简单可靠。     

Classification images for simple detection and discrimination tasks in correlated noise

We use the classification image technique to investigate the effect of white noise and various correlated Gaussian noise textures (low-pass, high-pass, and band-pass) on observer performance in detection and discrimination tasks. For these tasks, performance is generally enhanced by an observer's ability to "prewhiten" correlated noise as part of the formation of a decision variable. We find that observer efficiency in these tasks is well represented by the measured classification images and that human observers show strong evidence of adaptation to different correlated noise textures. This adaptation is captured in the frequency weighting of the classification images.     

Analytical-form model observers for decompressed images

We report a method for evaluating the performance of model observers for decompressed images in analytical form using compression noise statistics. It derives test statistics and detectabilities for the ideal observer, the nonprewhitening observer, the Hotelling observer, and the channelized Hotelling observer (CHO) on decompressed images. The derived CHO performance is validated using the Joint Photographic Experts Group (JPEG) compression algorithm. The validation results show that the derived CHO receiver operating characteristics (ROCs) and areas under ROC curves predict accurately their corresponding estimated values. These analytical-form quality measures of decompressed images provide a way to optimize compression algorithms analytically, subject to a model-observer performance criterion. They also provide a theoretical foundation for efforts to create a model observer for decompressed images.     

Experimental verifications of a joint transform correlator using compressed reference images

Single-target and multiple-target detections by using a joint transform correlator (JTC) with compressed reference images are experimentally verified. Two high-contrast images with different spatial-frequency content are used as test scenes. Although an effect of the additive noise on detection performance of the proposed correlator is more severe than that of the compression, the experimental results confirm the feasibility of implementing the JTC with compressed reference images.     

Quantitative assessment of image quality enhancement due to unsharp-mask processing in x-ray fluoroscopy

Spatial unsharp-mask processing and its variants are commonly used in x-ray radiography to enhance image contrast. We investigated the effect of three unsharp-masking filter kernels of different sizes on the detection of an advanced guidewire tip in simulated x-ray fluoroscopy image sequences. To isolate the effect of visual temporal processing, we repeated the experiments on single images. Filter gains were selected so that all three kernels increased the contrast of a 0.018-in. (0.457-mm) guidewire by a factor of 2 but had different effects on image noise and signal profiles. There was no statistically significant effect of unsharp masking on human-observer performance in single images. However, all three kernels significantly improved average performance in image sequences, and the guidewire contrast required for detection was reduced by 32%-40%. A prewhitening channelized observer model predicted the disparity between sequences and single images and fitted measurements at different kernel sizes well. A nonprewhitening observer model did not. We conclude that unsharp masking is a simple and effective method of improving guidewire visualization in fluoroscopically guided interventional procedures and that quantitative image quality studies are essential for evaluation of image-processing techniques in sequences such as x-ray fluoroscopy.