基于局部块效应的JPEG伪造图像的盲取证

本文基于对JPEG图像整体块效应的分析,定义了新的针对图像局部区域的块效应评价,并由此提出了一种有效的JPEG伪造图像盲取证方法。首先获得图像在水平方向和垂直方向的差分图像,然后将两个方向的差分图像分别进行特定大小的分块,再计算每个分块局部区域的块效应评价,根据待测图像不同区域局部块效应评价的明显差异检测出图像被篡改的位置。实验结果表明,该方法可以有效的检测出经过JPEG双压缩的伪造图像。      

基于多尺度稠密残差网络的JPEG压缩伪迹去除方法

JPEG在高压缩比的情况下,解压缩后的图像会产生块效应、边缘振荡效应和模糊,严重影响了图像的视觉效果。为了去除JPEG压缩伪迹,该文提出了多尺度稠密残差网络。首先把扩张卷积引入到残差网络的稠密块中,利用不同的扩张因子,使其形成多尺度稠密块;然后采用4个多尺度稠密块将网络设计成包含2条支路的结构,其中后一条支路用于补充前一条支路没有提取到的特征;最后采用残差学习的方法来提高网络的性能。为了提高网络的通用性,采用具有不同压缩质量因子的联合训练方式对网络进行训练,针对不同压缩质量因子训练出一个通用模型。经实验表明,该文方法不仅具有较高的JPEG压缩伪迹去除性能,且具有较强的泛化能力。     

基于多残差学习与注意力融合的中值滤波检测

针对小尺寸JPEG压缩图像携带有效信息较少、中值滤波痕迹不明显的问题,提出一种基于多残差学习与注意力融合的图像中值滤波检测算法。该算法将多个高通滤波器与注意力模块相结合,获取带权值的多残差特征图作为特征提取层的输入,特征提取层采用分组卷积形式,对输入的多残差特征图进行多尺度特征提取,融合不同尺度的特征信息,同时采用密集连接方式,每一层卷积的输入来自前面所有卷积层的输出和。实验结果表明,针对小尺寸JPEG压缩图像的中值滤波检测,本文算法比现有算法具有更高的检测精度,且能更有效地检测与定位局部篡改区域。     

Generalized block-lifting factorization of M-channel biorthogonal filter banks for lossy-to-lossless image coding

Generalized block-lifting factorization of M-channel (M > 2) biorthogonal filter banks (BOFBs) for lossy-to-lossless image coding is presented in this paper. Since the proposed block-lifting structure is more general than the conventional lifting factorizations and does NOT require many restrictions such as paraunitary, number of channels, and McMillan degree in each building block unlike the conventional lifting factorizations, its coding gain is higher than that of the previous methods. Several proposed BOFBs are designed and applied to image coding. Comparing the results with conventional lossy-to-lossless image coding structures, including the 5/3- and 9/7-tap discrete wavelet transforms in JPEG 2000 and a 4 × 8 hierarchical lapped biorthogonal transform in JPEG XR, the proposed BOFBs achieve better result in both objective measure and perceptual visual quality for the images with a lot of high-frequency components.     

基于多尺度级联网络的水下图像增强方法

针对水下图像由于光吸收、后向散射等因素导致的严重色偏、细节丢失等问题,该文提出一种基于多尺度级联网络的水下图像增强方法。针对单一网络特征利用不全面导致的图像梯度消失问题,该方法通过级联多尺度原始图像与相应的特征图像,以获得更优异的细节保持效果,并实现从较浅层到较深层快速预测残差的能力。此外,引入联合密集网络块和递归块,通过特征重用有效解决多尺度网络参数过多的问题。为有效解决单一损失造成的图像细节恢复不均的问题,提出Charbonnier和结构相似度 (SSIM) 联合损失函数。经仿真实验分析,所提网络在处理水下图像严重色偏、细节丢失等方面都取得了显著的效果。     

红外与可见光图像多特征自适应融合方法

由于成像机理不同,红外图像以像素分布表征典型目标,而可见光图像以边缘和梯度描述纹理细节,现有的融合方法不能依据源图像特征自适应变化,造成融合结果不能同时保留红外目标特征与可见光纹理细节。为此,本文提出红外与可见光图像多特征自适应融合方法。首先,构建了多尺度密集连接网络,可以有效聚合所有不同尺度不同层级的中间特征,利于增强特征提取和特征重构能力。其次,设计了多特征自适应损失函数,采用VGG-16网络提取源图像的多尺度特征,以像素强度和梯度为测量准则,以特征保留度计算特征权重系数。多特征自适应损失函数监督网络训练,可以均衡提取源图像各自的特征信息,从而获得更优的融合效果。公开数据集的实验结果表明,该方法在主、客观评价方面均优于其他典型方法。     

Detection and localization of forgery using statistics of DCT and Fourier components

In this paper, we present a comprehensive approach for investigating JPEG compressed test images, suspected of being tampered either by splicing or copy-move forgery (cmf). In JPEG compression, the image plane is divided into non-overlapping blocks of size 8 × 8 pixels. A unified approach based on block-processing of JPEG image is proposed to identify whether the image is authentic/forged and subsequently localize the tampered region in forged images. In the initial step, doubly stochastic model (dsm) of block-wise quantized discrete cosine transform (DCT) coefficients is exploited to segregate authentic and forged JPEG images from a standard dataset (CASIA). The scheme is capable of identifying both the types of forged images, spliced as well as copy-moved. Once the presence of tampering is detected, the next step is to localize the forged region according to the type of forgery. In case of spliced JPEG images, the tampered region is localized using block-wise correlation maps of dequantized DCT coefficients and its recompressed version at different quality factors. The scheme is able to identify the spliced region in images tampered by pasting uncompressed or JPEG image patch on a JPEG image or vice versa, with all possible combinations of quality factors. Alternatively, in the case of copy-move forgery, the duplication regions are identified using highly localized phase congruency features of each block. Experimental results are presented to consolidate the theoretical concept of the proposed technique and the performance is compared with the already existing state of art methods.     

Blocking effect reduction of JPEG images by signal adaptivefiltering

A postprocessing algorithm is proposed to reduce the blocking artifacts of Joint Photographic Experts Group (JPEG) decompressed images. The reconstructed images from JPEG compression produce noticeable image degradation near the block boundaries, in particular for highly compressed images, because each block is transformed and quantized independently. The blocking effects are classified into three types of noises in this paper: grid noise, staircase noise, and corner outlier. The proposed postprocessing algorithm, which consists of three stages, reduces these blocking artifacts efficiently. A comparison study between the proposed algorithm and other postprocessing algorithms is made by computer simulation with several JPEG images.     

基于自编码神经网络重构的车牌数字识别

提出了一种基于自编码神经网络重构的车牌数字识别方案.首先对车牌图像进行预处理,利用车牌字符的原图和Gabor特征作为自编码神经网络的输入进行识别实验.然后对每个车牌字符构造一个自编码神经网络,利用训练样本进行图像的重构训练,并根据训练得到的网络权值重构出训练样本集中的各个字符图像或特征.最后,将测试样本输入到每个自编码...     

块级篡改定位的JPEG图像脆弱水印

 JPEG是一种常见的图像格式,在JPEG图像中进行准确的篡改定位具有重要意义. 本文提出一种新的JPEG图像脆弱水印方案,将每个小块主要内容的Hash比特重新分组,并将每组的模2和作为水印信息. 也就是说每一个小块都对应多个水印比特,每个水印比特也对应多个小块. 载体图像的每个小块中仅嵌入于1比特水印,保证了良好的隐蔽性. 认证时依据图像内容与水印比特的整体匹配情况估计篡改率,再根据每个小块对应的水印信息的被破坏程度判别该小块是否曾被篡改. 理论分析和实验结果标明该方法可以在篡改区域小于1%的情况下准确地找到所有篡改小块.     

基于生成对抗网络和噪声水平估计的低剂量CT图像降噪方法

生成对抗网络(GAN)用于低剂量CT(LDCT)图像降噪具有一定的性能优势,成为近年CT图像降噪领域新的研究热点。不同剂量的LDCT图像中噪声和伪影分布的强度发生变化时,GAN网络降噪性能不稳定,网络泛化能力较差。为了克服这一缺陷,该文首先设计了一个编解码结构的噪声水平估计子网,用于生成不同剂量LDCT图像对应的噪声图,并用原始输入图像与之相减来初步抑制噪声;其次,在主干降噪网络中,采用GAN框架,并将生成器设计为多路编码的U-Net结构,通过博弈对抗实现网络结构优化,进一步抑制CT图像噪声;最后,设计了多种损失函数来约束不同功能模块的参数优化,进一步保障了LDCT图像降噪网络的性能。实验结果表明,与目前流行算法相比,所提出的降噪网络能够在保留LDCT图像原有重要信息的基础上,取得较好的降噪效果。     

Optimal quantisation strategy for DCT image compression

The authors present a strategy for generating optimal quantisation tables for use in JPEG image compression and its extension to general block sizes. Directly optimised quantisation tables were obtained by simulated annealing. A composite cost function minimised the RMS error between original and recovered images while keeping the compression ratio close to some desired value. Examination of these tables led to a simple model giving quantisation coefficients in terms of (x, y) position in the table and three model parameters. Annealing on the model parameters for several compressions yielded an expression for each parameter as a function of compression ratio. This approach was extended to general block sizes, and psychovisual evaluation determined the visually optimal block size for each compression ratio. The authors demonstrate significant improvements over JPEG coding due to the use of optimal quantisation rather than default tables. Use of general block size effectively extends the JPEG approach to higher compressions than are feasible with standard JPEG coding     

基于自注意力深度网络的图像超分辨率重建方法

针对现有图像超分辨重建方法难以充分重建图像的细节信息且易出现重建的图像缺乏层次的问题,提出一种基于自注意力深度网络的图像超分辨重建方法。以深度神经网络为基础,通过提取低分辨率图像特征,建立低分辨率图像特征到高分辨率图像特征的非线性映射,重建高分辨率图像。在进行非线性映射时,引入自注意力机制,获取图像中全部像素间的依赖关系,利用图像的全局特征指导图像重建,增强图像层次。在训练深度神经网络时,使用图像像素级损失和感知损失作为损失函数,以强化网络对图像细节信息的重建能力。在3类数据集上的对比测试结果表明,所提方法能够提升图像超分辨重建结果的细节信息,且重建图像的视觉效果更好。     

Non-reference image quality assessment based on deep clustering

Image quality assessment (IQA) is an indispensable technique in computer vision and pattern recognition Existing deep IQA methods have achieved remarkable performance. As far as we know, these deep learning-based IQA algorithms lack an adaptive features extraction mechanism toward input images with varying sizes and the stability in avoiding disturbance from data noises and model deviation. To solve these problems, we propose a non-reference IQA method by designing a novel unsupervised deep clustering framework, where a 13-layer network structure is proposed that upgrades the fully-connected layers to produce high-level features with adaptive sizes. Moreover, we add a contracted regular term with a contracted autoencoder into the clustering loss function to form a quality model reflecting the clustering structure. Compared to the other IQA algorithms, our model with simple structure exhibits more stable and robust performance by the initial configuration of network parameters during end-to-end training. The experimental results on the LIVE and CSIP databases have shown that our method not only performs better than the state-of-the-art IQA algorithms, but also has a simpler structure and better adaptability.     

密集残差网络红外图像超分辨率重构

单幅红外图像超分辨率重构算法作为红外图像分辨率提升应用的关键技术,近年来得到了广泛的研究。为了提高红外图像的分辨力,提出了一种基于残差密集对抗式生成网络的单幅红外图像分辨力提升方法。与以往基于对抗式生成网络的分辨力提升方法不同,本文方法的新颖性主要包含两个方面。首先,在网络架构方面进行改进,以提高性能。设计密集残差网络作为对抗式生成网络的生成网络,充分利用了低分辨率图像的有效特征。在生成网络中引入了一种连续内存机制,以利用密集的剩余块。其次,将Wasserstein-GAN作为损失函数,对判别网络模型进行修正,以达到稳定训练的目的。利用红外高分辨率图像数据集进行了大量的实验,结果表明,该方法在客观评价和主观评价方面均优于目前最新的方法。     

MLDNet: Multi-level dense network for multi-focus image fusion

In this study, we propose a new deep learning architecture named Multi-Level Dense Network (MLDNet) for multi-focus image fusion (MFIF). We introduce shallow and dense feature extraction in our feature extraction module to extract images features in a more robust way. In particular, we extracted the features from a mixture of many distributions from prior to the complex distribution through densely connected convolutional layers, then the extracted features are fused to form dense local feature maps. We added global feature fusion into the proposed architecture in order to merge the dense local feature maps of each source image into a fused image representation for the reconstruction of the final fused image. Our proposed MLDNet learns feature extraction, feature fusion and reconstruction within the same network to provide an end-to-end solution for MFIF. Experimental results demonstrate that our proposed method achieved significant performance against different state-of-the-art MFIF methods.     

基于自编码特征提取及弹性学习的手写数字识别

针对自编码算法提取输入特征能更好地发现样本间的相关性的优点,以自编码算法提取待识别样本特征作为多层前向网络的输入,以弹性BP算法训练网络,并用MNIST手写数字数据库样本测试。从正确率、拒识率、错误率和可靠率4项性能指标方面与逐像素方法进行了综合对比测试。研究表明,采用自编码特征提取、多层前向神经网络作为分类器以及弹性BP算法进行训练的手写数字识别,具有更快的收敛速度和更高的识别可靠率。     

基于坐标注意力机制融合的反无人机系统图像识别方法

反无人机系统是识别和打击“黑飞”无人机的有效手段,图像识别无人机是反无人机系统的关键之一。针对采集的无人机样本属于小样本、提取特征不够多,识别准确率不够高的问题,提出了一种基于迁移学习、密集卷积网络和坐标注意力机制融合的反无人机系统图像识别方法。首先,运用自制设备采集了多种无人机在不同背景下的图片,建立数据样本;其次,设计针对无人机小样本识别的基于迁移学习、坐标注意力机制和密集卷积网络融合的网络TL-CA4-DenseNet-121、基于通道注意力机制融合的网络TL-SE4-DenseNet-121等网络,运用设计的网络对小样本进行识别,并进行对比,然后分别进行了基于不同位置和不同个数的坐标注意力模块和通道注意力模块的网络识别实验;最后,将识别效果最优的网络与经典卷积神经网络模型进行对比实验。实验结果表明,提出的TL-CA4-DenseNet-121网络识别效果优于其他网络,识别的平均准确率为97.93%,F1-Score为0.9826,网络训练时间为6832 s。结果表明了该网络在识别小样本无人机方面的优越性和可行性。     

Image coding by block prediction of multiresolution subimages

The redundancy of the multiresolution representation has been clearly demonstrated in the case of fractal images, but it has not been fully recognized and exploited for general images. Fractal block coders have exploited the self-similarity among blocks in images. We devise an image coder in which the causal similarity among blocks of different subbands in a multiresolution decomposition of the image is exploited. In a pyramid subband decomposition, the image is decomposed into a set of subbands that are localized in scale, orientation, and space. The proposed coding scheme consists of predicting blocks in one subimage from blocks in lower resolution subbands with the same orientation. Although our prediction maps are of the same kind of those used in fractal block coders, which are based on an iterative mapping scheme, our coding technique does not impose any contractivity constraint on the block maps. This makes the decoding procedure very simple and allows a direct evaluation of the mean squared error (MSE) between the original and the reconstructed image at coding time. More importantly, we show that the subband pyramid acts as an automatic block classifier, thus making the block search simpler and the block matching more effective. These advantages are confirmed by the experimental results, which show that the performance of our scheme is superior for both visual quality and MSE to that obtainable with standard fractal block coders and also to that of other popular image coders such as JPEG.     

Multi-scale fusion residual encoder-decoder approach for low illumination image enhancement

The sensing light source of the line scan camera cannot be fully exposed in a low light environment due to the extremely small number of photons and high noise, which leads to a reduction in image quality. A multi-scale fusion residual encoder-decoder (FRED) was proposed to solve the problem. By directly learning the end-to-end mapping between light and dark images, FRED can enhance the image's brightness with the details and colors of the original image fully restored. A residual block (RB) was added to the network structure to increase feature diversity and speed up network training. Moreover, the addition of a dense context feature aggregation module (DCFAM) made up for the deficiency of spatial information in the deep network by aggregating the context's global multi-scale features. The experimental results show that the FRED is superior to most other algorithms in visual effect and quantitative evaluation of peak signa-to-noise ratio (PSNR) and structural similarity index measure (SSIM). For the factor that FRED can restore the brightness of images while representing the edge and color of the image effectively, a satisfactory visual quality is obtained under the enhancement of low-light.