Machine learning and directional switching median-based filter for highly corrupted images

In this paper, two-stage machine learning-based noise detection scheme has been proposed for identification of salt-and- pepper impulse noise which gives excellent detection results for highly corrupted images. In the first stage, a window of size $3\times 3$ is taken from image and some other features of this window are used as input to neural network. This scheme has distinction of having very low missed detection (MD) and false positives rates. In the second stage, decision tree-based algorithm (J48) is applied on some well-known statistical parameters to generate rules for noise detection. These noise detection methods give promising results for identification of noise from highly corrupted images. A modified version of switching median filter (directional weighted switching median filter) is proposed for noise removal. Performance of noise detector is measured using MD and false alarm FA. Filtering results are compared with state-of-the-art noise removal techniques in terms of peak signal-to-noise ratio and structural similarity index measure. Extensive experiments are performed to show that the proposed technique gives better results than state-of-the-art noise detection and filtering methods.     

SAR despeckling neural network with logarithmic convolutional product model

ABSTRACT

Synthetic aperture radar (SAR) images are inevitably contaminated by speckle noise due to its coherent imaging mechanism. Speckle noise obscures the intrinsic radar cross section (RCS) information in SAR images. This article proposes a novel deep neural network architecture specifically designed for despeckling purpose. It uses a convolutional neural network to extract image features and reconstruct a discrete RCS probability density function (PDF). It is trained by a hybrid loss function which measures the distance between the actual SAR image intensity PDF and the estimated one which is derived from convolution between the reconstructed RCS PDF and prior speckle PDF. The network can be trained by either purely simulated image patches or real SAR images. Experiment results on both simulated SAR images and real NASA/JPL AIRSAR images are used to test the performance, and the results show the efficacy of the proposed despeckling neural network compared with three state-of-the-art filters.     

Neuro‐filters used to enhance the image quality of video displays

Abstract— Digital images can be affected by external factors. There are many types of noise which affect digital images. Image filtration is a basic method used to suppress such hindrances. The disadvantage of most filtration methods and hardware filters created on their behalf is their inability to react to changes in the input signal. The structure of the filters used for image processing is similar to the structure of a bi‐dimensional neural‐network matrix. Investigations have shown that a system with serial‐parallel filters of any degree of complexity can be created on the basis of the neural‐network matrices. Each neural‐network matrix layer acts as a separate neuro‐filter which can be trained and adapted to changes in the characteristics of the images. The neural‐network matrices allow for the creation of various types of linear and nonlinear filters, as well as combinations on the basis of a uniform structure. It allows for the design of a universal hardware neuro‐filter structure that can perform as different types of filters by means of loading the connectors weight. In our paper, we consider the realization of neuro‐filters based on a neural‐network matrix, which allows the processing of both static and moving images and increases the image sharpness, suppresses the noise, and detects movable objects in the processed image.     

A new impulse detection and filtering method for removal of wide range impulse noises

A new impulse detection and filtering algorithm is proposed for restoration of images that are highly corrupted by impulse noise. It is based on the minimum absolute value of four convolutions obtained by one-dimensional Laplacian operators. The proposed algorithm can effectively remove the impulse noise with a wide range of noise density and produce better results in terms of the qualitative and quantitative measures of the images even at noise density as high as 90%. Extensive simulations show that the proposed algorithm provides better performance than many of the existing switching median filters in terms of noise suppression and detail preservation.     

局部特定空间关系统计特征的RVIN噪声检测器

目的 随机脉冲噪声（random-valued impulse noise，RVIN）检测器将局部图像统计值（local image statistics，LIS）作为图块中心像素点是否为噪声的判断依据，但LIS的描述能力较弱，在不同程度上制约了RVIN检测器的检测正确率，影响了后续开关型降噪模块的修复效果。为此，提出了一种基于局部特定空间关系统计特征的RVIN噪声检测器。方法 以局部中心像素点的8个邻域像素对数差值排序值（rank-ordered logarithmic difference，ROLD）并结合1个最小方向对数差值（minimum orientation logarithmic difference，MOLD）共9个反映局部特定空间关系的LIS统计值构成描述中心像素点是否为RVIN的噪声感知特征矢量，并通过在大量样本图块数据上提取的RVIN噪声感知特征矢量及其对应的噪声标签作为训练对（training pairs），训练获得一个基于多层感知网络（multi-layer perception，MLP）的RVIN噪声检测器。结果 对比实验从检测正确率和实际应用效果2个方面检验所提出的RVIN检测器的有效性，分别在10幅常用图像和50幅BSD （Berkeley segmentation data）纹理图像上进行测试，并与经典的脉冲噪声降噪算法中包含的噪声检测器以及MLPNNC （MLP neural network classifier）噪声检测器相比较，以漏检数、误检数和错检总数作为评价噪声检测正确率的指标。在常用图像集上本文所提RVIN检测器的漏检数和误检数较为平衡，在错检总数上排名处于所有对比算法中的前2名，为后续的降噪模块打下了很好的基础。在BSD纹理图像集上，将本文提出的RVIN检测器和GIRAF （generic iteratively reweighted annihilating filter）算法组合构成一种RVIN噪声降噪算法（proposed-GIRAF），proposed-GIRAF算法在50幅BSD图像上的峰值信噪比（peak signal-to-noise ratio，PSNR）均值在各个噪声比例下均取得了最优结果，与排名第2的对比算法相比，提升了0.471.96 dB。实验数据表明，所提出的RVIN噪声检测器的检测正确率优于现有的检测器，与修复算法联用后即可获得一种降噪效果更佳的开关型RVIN降噪算法。结论 本文提出的RVIN噪声检测器在各个噪声比例下具有鲁棒的预测准确性，配合GIRAF算法使用后，与经典的RVIN降噪算法相比，降噪效果最佳，具有很强的实用性。     

Fuzzy based Impulse Noise Reduction Method

In this paper, we propose an image filtering technique based on fuzzy logic control to remove impulse noise for low as well as highly corrupted images. The proposed method is based on noise detection, noise removal and edge preservation modules. The main advantage of the proposed technique over the other filtering techniques is its superior noise removal as well as detail preserving capability. Based on the criteria of peak-signal-to-noise-ratio (PSNR), mean square error (MSE), structural similarity index measure (SSIM) and subjective evaluation measure we have found experimentally that the proposed method provides much better performance than the state-of-the-art filters. To analyze the detail preservation capability of the proposed filter sensitivity analysis is performed by changing the detail preservation module to see its effects on the details (texture and edge information) of resultant image. This sensitivity analysis proves experimentally that significant image details have been preserved by the proposed method.     

Neuro-fuzzy synergism to the intelligent system for edge detection and enhancement

The paper presents a fuzzy neural network system for edge detection and enhancement. The system can both: (a) obtain edges and (b) enhance edges by recovering missing edges and eliminate false edges caused by noise. The research is comprised of three stages, namely, adaptive fuzzification which is employed to fuzzify the input patterns, edge detection by a three-layer feedforward fuzzy neural network, and edge enhancement by a modified Hopfield neural network. The typical sample patterns are first fuzzified. Then they are used to train the proposed fuzzy neural network. After that, the trained network is able to determine the edge elements with eight orientations. Pixels having high edge membership are traced for further processing. Based on constraint satisfaction and the competitive mechanism, interconnections among neurons are determined in the Hopfield neural network. A criterion is provided to find the final stable result that contains the enhanced edge measurement. The proposed neural networks are simulated on a SUN Sparc station. One hundred and twenty-three training samples are well chosen to cover all the edge and non-edge cases and the performance of the system will not be improved by adding more training samples. Test images are degraded by random noise up to 30% of the original images. Compared with standard edge detection operators and enhancement techniques, the proposed system based on the neuro-fuzzy synergism obtains very good results.     

基于GPN径向基神经网络的边缘检测方法

在基于神经网络的边缘检测模型中,大部分模型的检测效率不高,检测效果也有待提升.本文受人眼视觉系统特性的启发,提出了一种新的基于GPN （Gaussian Positive-Negative）径向基神经网络的边缘检测方法.首先,本文构造了一种新型的基于GPN径向基神经网络,将图像中经高斯滤波预处理后的每个像素点作为GPN径向基神经网络的中心点,并将其输入神经网络;然后,在每层之间使用卷积神经网络的部分特性进行处理,经过扩展层和隐层计算后输出结果;最后根据输出结果利用轮廓跟踪的方法将边缘提取出来.本文在检测效果以及效率这2个方面进行了相应的数值实验.针对合成图像以及部分灰度不均匀图像,相较于脉冲耦合神经网络模型、遗传神经网络模型以及卷积神经网络模型,本文模型在效率上得到了提升,且边缘的连通性更好.实验结果表明,本文提出的基于GPN径向基神经网络的边缘检测方法是一种新的、有效的边缘检测方法,比传统的神经网络边缘检测方法效率更高,且在检测效果上也有所提升.     

有效去除图像中脉冲噪声的新型滤波算法

提出一种基于局部极值噪声检测的迭代中值滤波算法．该算法集中了minrnax算法与PSM算法各自的优势，并将两种算法有机地结合起来．经过实验仿真并与其他滤波算法进行比较表明，该算法可以有效地去除图像中的脉冲噪声，尤其是在噪声密度非常大的情况下表现了很好的性能。     

基于改进伪中值滤波器的道路图像滤波算法*

针对已有的细胞神经网中值滤波器滤波时,收敛速度慢、稳定性不好以及滤波图像比较模糊的缺点,设计一种差值控制细胞神经网的改进伪中值滤波器。提出了改变取值空间、引入随机扰动、扩大中值滤波窗口尺度和引入Mask掩图的改进方法。实验结果表明：该算法具有去除各种强度脉冲随机噪声能力,又能保护图像细节信息,而且具有良好的实时性。     

Fuzzy clustering algorithms for unsupervised change detection in remote sensing images

In this paper, we propose a context-sensitive technique for unsupervised change detection in multitemporal remote sensing images. The technique is based on fuzzy clustering approach and takes care of spatial correlation between neighboring pixels of the difference image produced by comparing two images acquired on the same geographical area at different times. Since the ranges of pixel values of the difference image belonging to the two clusters (changed and unchanged) generally have overlap, fuzzy clustering techniques seem to be an appropriate and realistic choice to identify them (as we already know from pattern recognition literatures that fuzzy set can handle this type of situation very well). Two fuzzy clustering algorithms, namely fuzzy c-means (FCM) and Gustafson-Kessel clustering (GKC) algorithms have been used for this task in the proposed work. For clustering purpose various image features are extracted using the neighborhood information of pixels. Hybridization of FCM and GKC with two other optimization techniques, genetic algorithm (GA) and simulated annealing (SA), is made to further enhance the performance. To show the effectiveness of the proposed technique, experiments are conducted on two multispectral and multitemporal remote sensing images. A fuzzy cluster validity index (Xie-Beni) is used to quantitatively evaluate the performance. Results are compared with those of existing Markov random field (MRF) and neural network based algorithms and found to be superior. The proposed technique is less time consuming and unlike MRF does not require any a priori knowledge of distributions of changed and unchanged pixels.     

An image-based feature tracking algorithm for real-time measurement of clad height

This paper presents a novel algorithm for real-time detection of clad height in laser cladding which is known as a layered manufacturing technique. A real-time measurement of clad geometry is based on the use of a developed trinocular optical detector composed of three CCD cameras and the associated interference filters and lenses. The images grabbed by the trinocular optical detector are fed into an algorithm which combines an image-based tracking protocol and a recurrent neural network to extract the clad height in real-time. The image feature tracking strategy is a synergy between a simple image selecting protocol, a fuzzy thresholding technique, a boundary tracing method, a perspective transformation and an extraction of elliptical features of the projected melt pool’s images. The proposed algorithm and the trained network were utilized in the process resulting in excellent detection of the clad height at various working conditions in which SS303L was deposited on mild steel. It was concluded that the developed system can detect the clad height independent from clad paths with about 12% maximum error.     

基于递归滤波的边缘检测算法

针对包含不同程度噪声数字图像的边缘检测问题,提出了两种建立在房顶型模糊边缘模型基础上的平滑和边缘检测滤波器。首先用这些滤波算子以三阶递归的形式实现图像的平滑及梯度计算,再进行非极大值抑制及双阈值的边缘检测连接。实验结果表明,该方法得到的梯度图像均比Canny和Deriche滤波算子清晰,所得边缘图像也更加完整,检测时间也少于其他方法。     

采用训练策略实现的快速噪声水平估计

目的 大多数图像降噪算法都属于非盲降噪算法,其获得良好降噪性能的前提是能够准确地获知图像的噪声水平值。然而,现有的噪声水平估计（NLE）算法在噪声水平感知特征（NLAF）提取和噪声水平值映射两个核心模块中分别存在特征描述能力不足和预测准确性有待提高的问题。为此,提出了一种基于卷积神经网络（CNN）自动提取NLAF特征,并利用增强BP （back propagation）神经网络将其映射为相应噪声水平值的改进算法。方法 在训练阶段,首先通过训练卷积神经网络模型并以全连接层中若干与噪声水平值相关系数较高的输出值构成NLAF特征矢量;然后,在AdaBoost技术的支撑下,利用多个映射能力相对较弱的BP神经网络构建一个非线性映射能力更强的增强BP神经网络预测模型,将NLAF特征矢量直接映射为噪声水平值。在预测阶段,首先从给定噪声图像中随机选取若干个图块输入到卷积神经网络模型中,提取每个图块的若干维NLAF特征值后,利用预先训练的BP网络模型将其映射为对应的噪声水平值,然后以估计值的中值作为图像噪声水平值的最终估计结果。结果 对于具有不同噪声水平和内容结构的噪声图像,利用所提算法估计出的噪声水平值与真实值之间的估计误差小于0.5,均方根误差小于0.9,表现出良好的预测准确性和稳定性。此外,所提算法具有较高的执行效率,估计一幅512×512像素的图像的噪声水平值仅需约13.9 ms。结论 实验数据表明,所提算法在高、中、低各个噪声水平下都具有稳定的预测准确性和较高的执行效率,与现有的主流噪声水平估计算法相比综合性能更佳,可以很好地应用于要求噪声水平作为关键参数的实际应用中。     

Fuzzy clustering algorithms incorporating local information for change detection in remotely sensed images

In this paper we have used two fuzzy clustering algorithms, namely fuzzy c-means (FCM) and Gustafson–Kessel clustering (GKC) along with local information for unsupervised change detection in multitemporal remote sensing images. In conventional FCM and GKC no spatio-contextual information is taken into account and thus the result is not so much robust to small changes. Since the pixels are highly correlated with their neighbors in image space (spatial domain), incorporation of local information enhances the performance of the algorithms. In this work we have introduced a new technique for incorporation of local information. Change detection maps are obtained by separating the pixel-patterns of the difference image into two groups. Hybridization of FCM and GKC with two other optimization techniques, genetic algorithm (GA) and simulated annealing (SA), is made to further enhance the performance. To show the effectiveness of the proposed technique, experiments are conducted on two multispectral and multitemporal remote sensing images. Two fuzzy cluster validity measures (Xie–Beni and fuzzy hypervolume) have been used to quantitatively evaluate the performance. Results are compared with those of existing state of the art Markov random field (MRF) and neural network based algorithms and found to be superior. The proposed technique is less time consuming and unlike MRF does not require any a priori knowledge of distributions of changed and unchanged pixels.     

Medical image segmentation using a contextual-constraint-based Hopfield neural cube

Neural-network-based image techniques such as the Hopfield neural networks have been proposed as an alternative approach for image segmentation and have demonstrated benefits over traditional algorithms. However, due to its architecture limitation, image segmentation using traditional Hopfield neural networks results in the same function as thresholding of image histograms. With this technique high-level contextual information cannot be incorporated into the segmentation procedure. As a result, although the traditional Hopfield neural network was capable of segmenting noiseless images, it lacks the capability of noise robustness. In this paper, an innovative Hopfield neural network, called contextual-constraint-based Hopfield neural cube (CCBHNC) is proposed for image segmentation. The CCBHNC uses a three-dimensional architecture with pixel classification implemented on its third dimension. With the three-dimensional architecture, the network is capable of taking into account each pixel's feature and its surrounding contextual information. Besides the network architecture, the CCBHNC also differs from the original Hopfield neural network in that a competitive winner-take-all mechanism is imposed in the evolution of the network. The winner-take-all mechanism adeptly precludes the necessity of determining the values for the weighting factors for the hard constraints in the energy function in maintaining feasible results. The proposed CCBHNC approach for image segmentation has been compared with two existing methods. The simulation results indicate that CCBHNC can produce more continuous, and smoother images in comparison with the other methods.     

Video text detection and segmentation for optical character recognition

In this paper, we present approaches to detecting and segmenting text in videos. The proposed video-text-detection technique is capable of adaptively applying appropriate operators for video frames of different modalities by classifying the background complexities. Effective operators such as the repeated shifting operations are applied for the noise removal of images with high edge density. Meanwhile, a text-enhancement technique is used to highlight the text regions of low-contrast images. A coarse-to-fine projection technique is then employed to extract text lines from video frames. Experimental results indicate that the proposed text-detection approach is superior to the machine-learning-based (such as SVM and neural network), multiresolution-based, and DCT-based approaches in terms of detection and false-alarm rates. Besides text detection, a technique for text segmentation is also proposed based on adaptive thresholding. A commercial OCR package is then used to recognize the segmented foreground text. A satisfactory character-recognition rate is reported in our experiments.Published online: 14 December 2004     

Robust information gain based fuzzy c-means clustering and classification of carotid artery ultrasound images

In this paper, a robust method is proposed for segmentation of medical images by exploiting the concept of information gain. Medical images contain inherent noise due to imaging equipment, operating environment and patient movement during image acquisition. A robust medical image segmentation technique is thus inevitable for accurate results in subsequent stages. The clustering technique proposed in this work updates fuzzy membership values and cluster centroids based on information gain computed from the local neighborhood of a pixel. The proposed approach is less sensitive to noise and produces homogeneous clustering. Experiments are performed on medical and non-medical images and results are compared with state of the art segmentation approaches. Analysis of visual and quantitative results verifies that the proposed approach outperforms other techniques both on noisy and noise free images. Furthermore, the proposed technique is used to segment a dataset of 300 real carotid artery ultrasound images. A decision system for plaque detection in the carotid artery is then proposed. Intima media thickness (IMT) is measured from the segmented images produced by the proposed approach. A feature vector based on IMT values is constructed for making decision about the presence of plaque in carotid artery using probabilistic neural network (PNN). The proposed decision system detects plaque in carotid artery images with high accuracy. Finally, effect of the proposed segmentation technique has also been investigated on classification of carotid artery ultrasound images.     

采用多通道浅层CNN构建的多降噪器最优组合模型

现有的一致性神经网络(Consensus neural network, CsNet)利用凸优化和神经网络技术将多个降噪算法(降噪器)输出的图像进行加权组合(融合), 以获得更好的降噪效果, 但该优化模型在降噪效果和执行效率方面仍有较大改进空间. 为此, 提出一种基于轻量型多通道浅层卷积神经网络(Multi-channel shallow convolutional neural network, MSCNN)构建的多降噪器最优组合(Optimal combination of image denoisers, OCID)模型. 该模型采用多通道输入结构直接接收由多个降噪器输出的降噪图像, 并利用残差学习技术合并完成图像融合和图像质量提升两项任务. 具体使用时, 对于给定的一张噪声图像, 先用多个降噪器对其降噪, 并将降噪后图像输入OCID模型获得残差图像, 然后将多个降噪图像的均值图像与残差图像相减, 所得到图像作为优化组合后的降噪图像. 实验结果表明, 与CsNet组合模型相比, 网络结构更为简单的OCID模型以更小的计算代价获得了图像质量更高的降噪图像.     

An algorithmic approach to adaptive state filtering using recurrentneural networks

Practical algorithms are presented for adaptive state filtering in nonlinear dynamic systems when the state equations are unknown. The state equations are constructively approximated using neural networks. The algorithms presented are based on the two-step prediction-update approach of the Kalman filter. The proposed algorithms make minimal assumptions regarding the underlying nonlinear dynamics and their noise statistics. Non-adaptive and adaptive state filtering algorithms are presented with both off-line and online learning stages. The algorithms are implemented using feedforward and recurrent neural network and comparisons are presented. Furthermore, extended Kalman filters (EKFs) are developed and compared to the filter algorithms proposed. For one of the case studies, the EKF converges but results in higher state estimation errors that the equivalent neural filters. For another, more complex case study with unknown system dynamics and noise statistics, the developed EKFs do not converge. The off-line trained neural state filters converge quite rapidly and exhibit acceptable performance. Online training further enhances the estimation accuracy of the developed adaptive filters, effectively decoupling the eventual filter accuracy from the accuracy of the process model.