Removal of impulse noise in color images based on convolutional neural network

A new denoising framework based on deep convolutional neural network for suppressing impulse noise in color images is proposed in this paper. The proposed framework consists of two modules: noise detection and image reconstruction, both of which are implemented by a deep convolutional neural network. First, a noise classifier network is trained to detect random-valued impulse noise in a color image, which not only can detect the noisy color vector pixels but also can further identify the corrupted channels of each noisy color pixel. Then, a sparse clean color image is computed by replacing the values of noisy channels with 0 and keeping other noise-free channels unchanged. Finally, the sparse clean color image is fed to another denoiser network to reconstruct the denoised image. Experimental results show that the proposed denoiser outperforms other state-of-the-art methods clearly in both performance measure and visual evaluation.     

结合感知损失与双重对抗网络的低剂量CT图像去噪

低剂量计算机断层扫描（LDCT）成像技术在医学诊断中得到广泛应用,但其斑纹噪声和非平稳条纹伪影复杂,目前多数算法仅依靠推断条件后验概率来实现图像去噪,无法应对LDCT图像噪声复杂、数据量少、先验知识缺乏的问题。提出一种结合感知损失的双重对抗网络去噪算法,以实现LDCT图像复原。该算法包含一个去噪器和一个生成器,分别从图像去噪和噪声生成2个角度来建模干净-噪声图像对的联合分布,通过联合学习使得去噪器和生成器相互指导,从而充分学习数据中的噪声信息和清晰图像信息,且学习到的去噪器可以直接用于LDCT图像修复。考虑到通过感知损失学习语义特征差异可以使去噪结果保留更多的细节和边缘信息,提出一种掩膜自监督方法,针对CT图像域训练一个语义特征提取网络用于计算感知损失。实验结果表明,与BM3D、RED-CNN、WGAN-VGG等主流去噪算法相比,该算法可以有效抑制噪声并去除伪影,最大程度地保留边缘轮廓和纹理细节,产生更符合人眼视觉特性的去噪效果,与当下LDCT图像去噪性能较好的SACNN算法相比,所提算法的PSNR和SSIM指标分别提升1.26 dB和1.8%。     

A fusion estimation method based on fractional Fourier transform

Image denoising methods have different denoising performance in both spatial and transform domains, and each method has its relative advantages and inherent shortcomings compared with other methods. A very intuitive idea is to find that an effective fusion method that can combine with the advantages of different denoising methods. In this paper, we propose a novel fusion method based on the fractional Fourier transform and apply it to image denoising problem. Our method is mainly divided into three steps: Firstly, a pre-estimation is made by any two denoising method separately in the spatial domain. Secondly, using these two estimated results as well as their Fourier transform, twice Fourier transform and three times Fourier transform, we obtain a fused result in the fractional Fourier transform domain. Thirdly, the inverse fractional Fourier transform and the modulus operation are used to obtain the final fusion result. Obviously, this approach is the fusion method in four different domains. Experimental results on benchmark test images demonstrate that the proposed method outperforms state-of-the-art stand-alone methods as: BM3D, DDID, MLP, EPLL and also superior to the fusion methods such as classic wavelet fusion method, PCA fusion method and the state-of-the-art CIEM fusion method in terms of quantity value such as the peak signal to noise ratio (PSNR), the structural similarity (SSIM), and visual quality.     

结合第二代Bandelet变换分块的字典学习图像去噪算法

针对以往稀疏编码在图像去噪过程中存在的噪声残留和缺乏对图像的边缘与细节的本质特征的保护等问题,提出了一种结合第二代Bandelet变换分块的字典学习图像去噪算法,其更好地利用了图像的几何特性进行去噪。首先,通过第二代Bandelet变换可以灵活地根据图像几何流的正则性特征并能够自适应地获得图像的最稀疏表示来准确估计图像信息,并能自适应地选择最优的几何方向;然后,根据K-奇异值分解(K-Singular Value Decomposition,K-SVD)算法来训练学习字典;最后,通过四叉树分割对噪声图像进行自适应分块,从而去除噪声并保护图像的边缘与细节。实验结果表明,相比于其他学习字典,所提算法能更有效地保留图像的边缘特征与图像的精细结构。     

一种基于非局部思想的改进图像降噪算法

数字图像因噪声的影响会严重降低其视觉效果,图像降噪算法的研究是数字图像处理领域的一个重要研究方向。本文在基于稀疏和冗余字典的图像降噪算法基础上,提出了一种基于非局部思想的改进图像降噪算法。与传统的基于稀疏表达的图像降噪算法KSVD相比,本文算法增加了一个相似块聚合的过程,使得学习的字典更小且更准确。利用自然图像包含很多的自相似,相似样本聚合学习出的字典比传统KSVD算法能更准确更稀疏的表示样本。稀疏度的提高使得重建后的信号更加的准确,适应性更好。实验证明本文算法取得了更好的视觉效果。     

Multi-channel satellite cloud image fusion in the tetrolet transform domain

A novel multi-channel satellite cloud image fusion algorithm constructed in the tetrolet transform domain is proposed. Tetrolet is successfully applied in image denoising, image sparse representation, and image restoration. In this paper, tetrolet transform was introduced into the field of satellite cloud image fusion since its sparse degree is high. Tetrolet can describe the geometric structure feature of the satellite cloud image very well. First, tetrolet transform must be implemented into the multi-channel satellite cloud images to obtain low- and high-frequency coefficients and corresponding covering distribution values. Then, a Laplacian pyramid algorithm must be used to decompose the low-frequency portion in the tetrolet domain by averaging the values of its top layer and taking the maximum absolute values of the other layers. While reconstruction is implemented in this stage, the algorithm takes the maximum standard deviation of the high-frequency parts for each block in the tetrolet domain. Last, an inverse tetrolet transform must be used to obtain the final fused image. This paper compares the proposed image fusion algorithm to three similar image fusion algorithms: the curvelet image fusion algorithm, the non-subsampled contourlet transform (NSCT) image fusion algorithm, and the tetrolet image fusion algorithm. Mutual information, joint entropy, mean structural similarity (MSSIM), standard deviation, and average relative deviation are used as objective criteria to evaluate the quality of the fused images. In order to verify the efficiency of the proposed algorithm, the fusion cloud image is used to determine the centre location of eye and non-eye typhoons. Experimental results show that the proposed algorithm performs well when fusing the information in multi-channel satellite cloud images and improves the precision of locating the typhoon’s centre. The proposed algorithm’s comprehensive performance is superior to similar image fusion algorithms.     

基于偏微分方程的医学超声图像降噪研究

提出一种将图像分解和几何分析相结合的算法去除超声图像中的斑点噪声。针对超声图像的斑点噪声为乘性噪声特性,将经典的ROF图像分解模型引入到适合于受乘性噪声污染的图像分解。超声图像经模型分解为轮廓部分、细节部分和噪声部分,然后对分解后的差值图像进行Ridgelet降噪,由于Ridgelet降噪克服传统小波分析方向性上的不敏感的缺点,能很好地保持图像边缘。处理后得到的图像无论是在斑点噪声去除、细节保护方面都优于传统的非线性滤波器和小波分析方法。实验表明,算法是完全可行和有效的。     

单通道双谱夜视系统中的图像修复及消噪技术

基于单通道双谱夜视系统中对图像修复的需要，提出了一种基于水平集的图像修复及消噪技术。在阐述单通道双谱夜视系统工作原理的基础上，结合系统的实时性要求及待修复微光条纹图像的特点，设计了处理速度高的水平集修复算法；考虑微光图像的噪声对修复结果的影响，在修复的同时加入了热传导滤波方法。实验结果表明，该算法能够对微光条纹图像进行实时且有效的修复。     

Nonlocal Image and Movie Denoising

Neighborhood filters are nonlocal image and movie filters which reduce the noise by averaging similar pixels. The first object of the paper is to present a unified theory of these filters and reliable criteria to compare them to other filter classes. A CCD noise model will be presented justifying the involvement of neighborhood filters. A classification of neighborhood filters will be proposed, including classical image and movie denoising methods and discussing further a recently introduced neighborhood filter, NL-means. In order to compare denoising methods three principles will be discussed. The first principle, “method noise”, specifies that only noise must be removed from an image. A second principle will be introduced, “noise to noise”, according to which a denoising method must transform a white noise into a white noise. Contrarily to “method noise”, this principle, which characterizes artifact-free methods, eliminates any subjectivity and can be checked by mathematical arguments and Fourier analysis. “Noise to noise” will be proven to rule out most denoising methods, with the exception of neighborhood filters. This is why a third and new comparison principle, the “statistical optimality”, is needed and will be introduced to compare the performance of all neighborhood filters. The three principles will be applied to compare ten different image and movie denoising methods. It will be first shown that only wavelet thresholding methods and NL-means give an acceptable method noise. Second, that neighborhood filters are the only ones to satisfy the “noise to noise” principle. Third, that among them NL-means is closest to statistical optimality. A particular attention will be paid to the application of the statistical optimality criterion for movie denoising methods. It will be pointed out that current movie denoising methods are motion compensated neighborhood filters. This amounts to say that they are neighborhood filters and that the ideal neighborhood of a pixel is its trajectory. Unfortunately the aperture problem makes it impossible to estimate ground true trajectories. It will be demonstrated that computing trajectories and restricting the neighborhood to them is harmful for denoising purposes and that space-time NL-means preserves more movie details.     

Neural Temporal Adaptive Sampling and Denoising

Despite recent advances in Monte Carlo path tracing at interactive rates, denoised image sequences generated with few samples per-pixel often yield temporally unstable results and loss of high-frequency details. We present a novel adaptive rendering method that increases temporal stability and image fidelity of low sample count path tracing by distributing samples via spatio-temporal joint optimization of sampling and denoising. Adding temporal optimization to the sample predictor enables it to learn spatio-temporal sampling strategies such as placing more samples in disoccluded regions, tracking specular highlights, etc; adding temporal feedback to the denoiser boosts the effective input sample count and increases temporal stability. The temporal approach also allows us to remove the initial uniform sampling step typically present in adaptive sampling algorithms. The sample predictor and denoiser are deep neural networks that we co-train end-to-end over multiple consecutive frames. Our approach is scalable, allowing trade-off between quality and performance, and runs at near real-time rates while achieving significantly better image quality and temporal stability than previous methods.     

基于软门限去噪的图象压缩编码研究

在详细地分析了Donoho提出的子波域软限去噪方法的基础上,给出了含噪图象信号噪声水平的估计及门限值随尺度变化的规律。采用可分离的二维子波滤波器,方便地将Donoho的软门限去噪方法应用于图象信号处理,从而对含噪图象,在去除噪声的同时,又最大限度地进行了压缩。针对含噪的自然景物图象和合成孔径雷达图象的不同特点,分别提出了这在图象的压缩方案。对于SAR图象的压缩编码,通过一个自然对数变换,使得乘性噪声转变为适于软门限去噪的加性噪声。模拟结果显示,用软门限方法处理的解压缩图象比硬门限方法具有更好的视觉质量,因而该方法是解决含噪图象压缩编码的有效技术。     

一种新的非下采样Contourlet域图像去噪算法

作为新型高维奇异性分析工具,非下采样轮廓(Nonsubsampled Contourlet)变换不仅克服了小波(Wavelet)变换的非奇异性最优基缺点,而且提供了优于轮廓(Contourlet)变换的平移不变性.以性能优越的非下采样轮廓变换为基础,提出了一种新的图像去噪方法.该方法首先对图像进行非下采样轮廓变换,以得到不同尺度、不同方向上的变换系数;然后结合噪声分布特点确定多尺度阈值,并依此阚值对高频系数进行去噪处理;最后对去噪处理后的变换系数进行反变换,以得到去噪图像.仿真实验结果表明,该方法不仅拥有较强的抑制噪声的能力,而且具有较好的边缘保护能力,同时消除了图像边缘附近的伪吉布斯(Gibbs)现象,整体性能优于小波变换图像去噪和轮廓变换图像去噪方法.     

Multi-focus image fusion using PCNN

This paper proposes a new method for multi-focus image fusion based on dual-channel pulse coupled neural networks (dual-channel PCNN). Compared with previous methods, our method does not decompose the input source images and need not employ more PCNNs or other algorithms such as DWT. This method employs the dual-channel PCNN to implement multi-focus image fusion. Two parallel source images are directly input into PCNN. Meanwhile focus measure is carried out for source images. According to results of focus measure, weighted coefficients are automatically adjusted. The rule of auto-adjusting depends on the specific transformation. Input images are combined in the dual-channel PCNN. Four group experiments are designed to testify the performance of the proposed method. Several existing methods are compared with our method. Experimental results show our presented method outperforms existing methods, in both visual effect and objective evaluation criteria. Finally, some practical applications are given further.     

基于边缘检测的图象小波阈值去噪方法

边缘特征是图象最为有用的高频信息，因此，在图象去噪的同时，尽量保留图象的边缘特征，应是图象去噪首要顾及的问题。基于这一思想，提出了基于边缘检测的图象小波阈值去噪方法。该方法在去噪之前，先通过小波边缘检测方法确定哪些小波系数是图象的边缘特征，这些小波系数将不受阈值去噪的影响，因此，可以只是根据噪声方差来设置去噪的阈值，而不必担心损害图象的边缘特征。理论分析和实验结果都表明，与普通的小波阈值去噪方法相比，该方法不但可以保持图象的边缘信息，而且能提高去噪后图象的峰值信噪比1-2dB。要做到既去除图象噪声，又不模糊图象边缘特征是很困难的。该方法把去噪和边缘检测结合起来，在一定程度上解决了这种两难的问题。     

基于U-Net融合的保留纹理的图像去噪方法

图像的噪声阻碍了高级视觉任务对图像的理解,且去除图像的噪声是一个具有挑战性的任务.现有的基于卷积神经网络的图像去噪方法在去除噪声的同时,对图像纹理会引入一定程度的破坏,导致去噪后图像无法保留图像的纹理.为了解决这个问题,本文提出一种用二分支U-Net网络来融合特征和保留纹理的图像去噪方法.首先选取一种去噪方法的两个不同去噪参数的预训练模型分别得到同一张噪声图像的不同去噪结果,其中一个结果中去噪效果比纹理保留效果好,另一个结果中纹理保留比去噪效果好.然后将这两个去噪图像作为卷积神经网络的输入,利用两个编码器分别提取图像的特征,并同时放入融合模块融合图像的特征,最后利用解码器重建出无噪声图像.实验结果表明,与现有的方法相比本文的方法更有效,在去除噪声的同时能保留更多的图像纹理信息.     

基于TMS320C6x的双波段图象高速融合系统

对自主研制的，以TMS320C6201数字信号处理器（DSP）为核心处理器的高速图象融合系统的设计与实现方案进行了完整描述，并着重讨论了双通道数字图象融合处理硬件系统设计中的特殊问题，由于采用了最新的高性能DSP以及硬件结构优化设计，该系统可以灵活地应用多种融合算法来实现可见光-长波红外双通道数字图象的实时或准实时融合处理，并具有手动像素平移配准功能，可以较好地解决多尺度图象融合算法的大数据量计算处理与硬件系统实时性要求之间的矛盾，为实用化的多通道实时图象融合处理机的研制工作奠定了良好的技术基础。     

An approach based on hybrid genetic algorithm applied to image denoising problem

An approach based on hybrid genetic algorithm (HGA) is proposed for image denoising. In this problem, a digital image corrupted by a noise level must be recovered without losing important features such as edges, corners and texture. The HGA introduces a combination of genetic algorithm (GA) with image denoising methods. During the evolutionary process, this approach applies some state-of-the-art denoising methods and filtering techniques, respectively, as local search and mutation operators. A set of digital images, commonly used by the scientific community as benchmark, is contaminated by different levels of additive Gaussian noise. Another set composed of some Satellite Aperture Radar (SAR) images, corrupted with a multiplicative speckle noise, is also used during the tests. First, the computational tests evaluate several alternative designs from the proposed HGA. Next, our approach is compared against literature methods on the two mentioned sets of images. The HGA performance is competitive for the majority of the reported results, outperforming several state-of-the-art methods for images with high levels of noise.     

结合多分辨率与块分割的不同聚焦图像融合

提出了一种结合多分辨率与图像块分割的图像融合新方法。该方法除将多分辨率融合图像块作为可能位于源图像的模糊区域与清晰区域交界处位置的最终融合图像块之外,还将多分辨率图像融合方法获取的融合图像与源图像进行分块比较,选用与多分辨率融合图像块相似的源图像块作为最终融合图像块。实验结果表明,该方法能有效提高常用的多分辨率图像融合方法的融合效果。     

A new wavelet-based fuzzy single and multi-channel image denoising

In this paper, we propose a new wavelet shrinkage algorithm based on fuzzy logic. In particular, intra-scale dependency within wavelet coefficients is modeled using a fuzzy feature. This feature space distinguishes between important coefficients, which belong to image discontinuity and noisy coefficients. We use this fuzzy feature for enhancing wavelet coefficients' information in the shrinkage step. Then a fuzzy membership function shrinks wavelet coefficients based on the fuzzy feature. In addition, we extend our noise reduction algorithm for multi-channel images. We use inter-relation between different channels as a fuzzy feature for improving the denoising performance compared to denoising each channel, separately. We examine our image denoising algorithm in the dual-tree discrete wavelet transform, which is the new shiftable and modified version of discrete wavelet transform. Extensive comparisons with the state-of-the-art image denoising algorithm indicate that our image denoising algorithm has a better performance in noise suppression and edge preservation.