A Degradation Type Adaptive and Deep CNN-Based Image Classification Model for Degraded Images

Deep Convolutional Neural Networks(CNNs)have achieved high accuracy in image classification tasks,however,most existing models are trained on high-quality images that are not subject to image degradation.In practice,images are often affected by various types of degradation which can significantly impact the performance of CNNs.In this work,we investigate the influence of image degradation on three typical image classification CNNs and propose a Degradation Type Adaptive Image Classification Model(DTA-ICM)to improve the existing CNNs’classification accuracy on degraded images.The proposed DTA-ICM comprises two key components:a Degradation Type Predictor(DTP)and a Degradation Type Specified Image Classifier(DTS-IC)set,which is trained on existing CNNs for specified types of degradation.The DTP predicts the degradation type of a test image,and the corresponding DTS-IC is then selected to classify the image.We evaluate the performance of both the proposed DTP and the DTA-ICMon the Caltech 101 database.The experimental results demonstrate that the proposed DTP achieves an average accuracy of 99.70%.Moreover,the proposed DTA-ICM,based on AlexNet,VGG19,and ResNet152,exhibits an average accuracy improvement of 20.63%,18.22%,and 12.9%,respectively,compared with the original CNNs in classifying degraded images.It suggests that the proposed DTA-ICM can effectively improve the classification performance of existing CNNs on degraded images,which has important practical implications.

Novel Adaptive Binarization Method for Degraded Document Images

Achieving a good recognition rate for degraded document images is difficult as degraded document images suffer from low contrast,bleedthrough,and nonuniform illumination effects.Unlike the existing baseline thresholding techniques that use fixed thresholds and windows,the proposed method introduces a concept for obtaining dynamic windows according to the image content to achieve better binarization.To enhance a low-contrast image,we proposed a new mean histogram stretching method for suppressing noisy pixels in the background and,simultaneously,increasing pixel contrast at edges or near edges,which results in an enhanced image.For the enhanced image,we propose a new method for deriving adaptive local thresholds for dynamic windows.The dynamic window is derived by exploiting the advantage of Otsu thresholding.To assess the performance of the proposed method,we have used standard databases,namely,document image binarization contest(DIBCO),for experimentation.The comparative study on well-known existing methods indicates that the proposed method outperforms the existing methods in terms of quality and recognition rate.

Visibility Enhancement of Scene Images Degraded by Foggy Weather Condition: An Application to Video Surveillance

:In recent years,video surveillance application played a significant role in our daily lives.Images taken during foggy and haze weather conditions for video surveillance application lose their authenticity and hence reduces the visibility.The reason behind visibility enhancement of foggy and haze images is to help numerous computer and machine vision applications such as satellite imagery,object detection,target killing,and surveillance.To remove fog and enhance visibility,a number of visibility enhancement algorithms and methods have been proposed in the past.However,these techniques suffer from several limitations that place strong obstacles to the real world outdoor computer vision applications.The existing techniques do not perform well when images contain heavy fog,large white region and strong atmospheric light.This research work proposed a new framework to defog and dehaze the image in order to enhance the visibility of foggy and haze images.The proposed framework is based on a Conditional generative adversarial network(CGAN)with two networks;generator and discriminator,each having distinct properties.The generator network generates fog-free images from foggy images and discriminator network distinguishes between the restored image and the original fog-free image.Experiments are conducted on FRIDA dataset and haze images.To assess the performance of the proposed method on fog dataset,we use PSNR and SSIM,and for Haze dataset use e,r−,andσas performance metrics.Experimental results shows that the proposed method achieved higher values of PSNR and SSIM which is 18.23,0.823 and lower values produced by the compared method which are 13.94,0.791 and so on.Experimental results demonstrated that the proposed framework Has removed fog and enhanced the visibility of foggy and hazy images.

Research on a novel restoration algorithm of turbulence-degraded images with alternant iterations

A new restoration algorithm based on double loops and alternant iterations is proposed to restore the object image effectively from a few frames of turbulence-degraded images, Based on the double loops, the iterative relations for estimating the turbulent point spread function PSF and object image alternately are derived. The restoration experiments have been made on computers, showing that the proposed algorithm can obtain the optimal estimations of the object and the point spread function, with the feasibility and practicality of the proposed algorithm being convincing.

Multiresolution Fusion of Remote Sensing Images Based on Resolution Degradation Model

A new method based on resolution degradation model is proposed to improve both spatial and spectral quality of the synthetic images. Some ETM+ panchromatic and multispectral images are used to assess the new method. Its spatial and spectral effects are evaluated by qualitative and quantitative measures and the results are compared with those of IHS, PCA, Brovey, OWT(Orthogonal Wavelet Transform) and RWT(Redundant Wavelet Transform). The results show that the new method can keep almost the same spatial resolution as the panchromatic images, and the spectral effect of the new method is as good as those of wavelet-based methods.

Determination of the Early Time of Death by Computerized Image Analysis of DNA Degradation: Which Is the Best Quantitative Indicator of DNA Degradation?

This study evaluated the correlation between DNA degradation of the splenic lymphocytes and the early time of death, examined the early time of death by computerized image analysis technique （CIAT） and identified the best parameter that quantitatively reflects the DNA degradation. The spleen tissues from 34 SD rats were collected, subjected to cell smearing every 2 h within the first 36 h after death, stained by Feulgen-Van＇s staining, three indices reflecting DNA content in splenic lymphocytes, including integral optical density （IOD）, average optical density （AOD）, average gray scale （AG） were measured by the image analysis. Our results showed that IOD and AOD decreased and AG increased over time within the first 36 h. A stepwise linear regression analysis showed that only AG was fitted. A correlation between the postmortem interval （PMI） and AG was identified and the corresponding regression equation was obtained. Our study suggests that CIAT is a useful and promising tool for the estimation of early PMI with good objectivity and reproducibility, and AG is a more effective and better quantitative indicator for the estimation of PMI within the first 36 h after death in rats.

Evaluation of influences of frequency and amplitude on image degradation caused by satellite vibrations

Satellite vibrations during exposure will lead to pixel aliasing of remote sensors, resulting in the deterioration of image quality. In this paper, we expose the problem and discuss the characteristics of satellite vibrations, and then present a pixel mixing model. The idea of mean mixing ratio （MMR） is proposed. MMR computations for different frequencies are implemented. In the mixing model, a coefficient matrix is introduced to estimate each mixed pixel. Thus, the simulation of degraded image can be performed when the vibration attitudes are known. The computation of MMR takes into considera- tion the influences of various frequencies and amplitudes. Therefore, the roles of these parameters played in the degradation progress are identified. Computations show that under the same vibration amplitude, the influence of vibrations fluctuates with the variation of frequency. The fluctuation becomes smaller as the frequency rises. Two kinds of vibration imaging experiments are performed： different amplitudes with the same frequency and different frequencies with the same amplitude. Results are found to be in very good agreement with the theoretical results. MMR has a better description of image quality than modulation transfer function （MTF）. The influence of vibrations is determined mainly by the amplitude rather than the frequency. The influence of vibrations on image quality becomes gradually stable with the increase of frequency.

A Hybrid-Binarization Approach for Degraded Document Enhancement

Images get degraded because of unbalanced enlightenment including text-smearing, ink-bleeding, degradation of ink over time, manuscript characters from background coming out and blended with the characters of the main side etc. So, degraded-document enhancement is a challenging issue. In recent years, several binarization approaches are proposed to enhance these images. These techniques have focused on finding a suitable global threshold value or a local threshold value for every region to eliminate the degradations. A hybrid approach can be a good solution to deal with all these matters together. This paper proposes a hybrid approach of binarization for degraded documents to produce better quality result. Then, the performance of the proposed technique is evaluated using DIBCO 2010 to DIBCO 2018 databases and compared with the existing methods which confirmed that the proposed method is robust, efficient. Finally, a direction towards future works and challenges is stated.

基于改进U-Net的水下图像增强算法

针对水下退化图像存在颜色失真、模糊雾化、对比度低等问题,提出了一种新的基于改进U-Net的水下图像增强算法。设计一种新的残差注意力结构和边缘检测模块并将其引入到U-Net网络中,构建改进后的水下图像增强算法。实验结果表明,本文提出的算法在校正水下色偏和增强对比度方面均得到了很好的效果,IE值较原始图像平均提高了14.2%,UCIQE值较原始图像平均提高了24%。消融实验结果表明,本文提出的残差注意力结构、边缘检测模块和损失函数均对水下图像增强起到了积极的效果。

基于深度学习的红外成像退化模型辨识及超分辨率成像方法

红外成像系统由于制造工艺和成本制约,分辨率仍然较低。图像超分辨率重建技术是提高图像分辨率的有效方法,获得了广泛研究,并在仿真图像上获得了很好的效果,但应用于实际图像时效果不甚理想,主要原因是实际成像退化更加复杂,包括红外光学系统像差和装配误差引起的空间非一致模糊,以及受工作温度影响导致的模糊核变化。针对上述问题,提出一种基于深度学习的红外成像退化模型辨识方法和基于退化模型约束的超分辨率重建方法,通过在不同工作温度下采集标定靶标图像,标定不同工作温度、不同空间位置的模糊核;采用卷积神经网络建立成像退化模型,并利用定标数据进行模型参数求解,为超分辨率重建提供更多先验信息;设计迭代超分辨率重建网络,交替进行退化参数估计和超分辨率重建,经过多次迭代逐步提高重建效果。实验结果表明,采用卷积神经网络求解的成像退化模型可准确描述模糊核变化规律,基于退化模型约束和退化参数在线学习的超分辨率重建方法可显著提高红外超分辨率成像的效果,具有较高的工程应用价值。

面向真实场景的单帧红外图像超分辨率重建

现有的红外图像超分辨率重建方法主要依赖实验数据进行设计,但在面对真实环境中的复杂退化情况时,它们往往无法稳定地表现。针对这一挑战,本文提出了一种基于深度学习的新颖方法,专门针对真实场景下的红外图像超分辨率重建,构建了一个模拟真实场景下红外图像退化的模型,并提出了一个融合通道注意力与密集连接的网络结构。该结构旨在增强特征提取和图像重建能力,从而有效地提升真实场景下低分辨率红外图像的空间分辨率。通过一系列消融实验和与现有超分辨率方法的对比实验,本文方法展现了其在真实场景下红外图像处理中的有效性和优越性。实验结果显示,本文方法能够生成更锐利的边缘,并有效地消除噪声和模糊,从而显著提高图像的视觉质量。

基于混合域注意力机制的神经网络反演大气湍流强度

本文提出了一种基于混合域注意力机制的神经网络方法反演大气湍流强度。神经网络的输入为不同大气湍流强度下的退化图像,输出为表征大气湍流强度的折射率结构常数。混合域注意力机制由空间域和通道域双重注意力机制组成,其中空间域注意力机制用于增强退化图像中受湍流影响的区域特征,通道域注意力机制用于增强由湍流引起的颜色和纹理特征。在网络训练阶段,引入的混合域注意力机制让神经网络更专注于退化图像中与大气湍流强度相关的特征,提高了模型的精度。数值实验结果表明,本文提出的方法能够较准确地实现大气湍流强度反演。

用于视觉传达效果优化的退化图像复原仿真

大气湍流与外界噪声的干扰,会造成图像出现退化现象,影响图像质量。为此提出考虑视觉传达效果的多帧退化图像复原算法。图像复原前构建出图像退化模型,明确图像退化规律,利用小波变换对多帧退化图像实施去噪处理。计算因大气湍流导致的相位扰动的波前图像,根据相位偏差,生成Zernike波前图像重建模型,多次迭代后完成PDF的波前重建,利用代价函数生成多帧退化图像复原的迭代公式,依据判定条件实现多帧退化图像复原。实验结果表明,所提算法的图像复原效果较好,复原效率高,具有优异的图像去噪效果。

基于生成对抗网络的轻量级图像盲超分辨率网络

针对图像盲超分辨率网络计算参数多、模型庞大的问题,对快速且节省内存的轻量级图像非盲超分辨率网络(fast and memory-efficient image super resulotion network,FMEN)进行改进,提出了一种轻量级的快速且节省内存的图像盲超分辨率网络(fast and memory-efficient image blind super resulotion network,FMEBN)。首先,通过图像退化模块模拟部分真实世界退化空间,使用退化预测模块预测低分辨率(low resolution,LR)图像的退化参数;然后,为能有效利用退化先验信息指导并约束网络进行重建,使用动态卷积对原网络特征提取、重建模块、高频注意力块(high frequency attention block,HFAB)结构进行改进;最后,使用生成对抗网络(generative adversarial network,GAN)对FMEN训练策略与损失函数进行优化,减小真实数据与生成数据的差异,生成更加真实、清晰的纹理、轮廓。实验结果表明,在合成图像数据集和真实图像数据集RealWorld-38上,该算法有较好的重建精度与视觉效果,模型大小12 MB,可以满足图像盲超分辨率网络的轻量级需求。

变幅反复荷载作用下变形钢筋与混凝土间的黏结退化

反复荷载下钢筋与混凝土间的黏结滑移本构模型是震后钢筋混凝土结构整体性能评估及加固修复的重要依据。由于黏结应力主要由变形钢筋与混凝土间的机械咬合力所组成,黏结区域侧向约束状态会对其黏结性能产生影响。基于此,研究了在侧向压力作用下变幅反复荷载对钢筋与混凝土黏结性能的影响,定量分析了黏结参数随变幅反复荷载的退化规律,并利用数字图像技术分析了钢筋与混凝土间的黏结破坏机理。研究表明,在混凝土保护层充分约束条件下,钢筋与混凝土间的黏结退化主要与反复荷载控制位移有关,而侧向压力对其影响可忽略。

基于生成逆推的大气湍流退化图像复原方法

大气湍流是影响远距离成像质量的重要因素。虽然已有的深度学习模型能够较好地抑制大气湍流引起的图像像素几何位移与空间模糊,但是这些模型需要大量的参数和计算量。为了解决该问题,提出了一种轻量化的基于生成逆推的大气湍流退化图像复原模型,该模型包含了去模糊、去偏移和湍流再生成三个核心模块。其中,去模糊模块通过高维特征映射块、细节特征抽取块和特征补充块,抑制湍流引起的图像模糊;去偏移模块通过两层卷积,补偿湍流引起的像素位移;湍流再生成模块通过卷积等操作再次生成湍流退化图像。在去模糊模块中,设计了基于注意力的特征补充模块,该模块融合了通道注意力机制与空间混合注意力机制,能在训练过程中关注图像中的重要细节信息。在公开的Heat Chamber与自建的Helen两个数据集上,所提模型分别取得了19.94 dB、23.51 dB的峰值信噪比和0.688 2、0.752 1的结构相似性。在达到当前最佳SOTA方法性能的同时,参数量与计算量有所减少。实验结果表明,该方法对大气湍流退化图像复原有良好的效果。

基于真实退化估计与高频引导的内窥镜图像超分辨率重建

内窥镜是诊断人体器官疾病的重要医疗设备,然而受人体内腔环境影响,内窥镜图像分辨率一般较低,需对其进行超分辨处理.目前多数基于深度学习的超分辨算法直接使用双三次插值下采样从高质量图像中获取低分辨率(Low-resolution, LR)图像以进行配对训练,此种方式会导致纹理细节丢失,不适用于医学图像.为解决该问题,针对医学内窥镜图像开发了一种新颖的退化框架,首先从真实低质量内窥镜图像中提取丰富多样的真实模糊核与噪声模式,之后提出一种退化注入算法,利用提取的真实模糊核与噪声将高分辨率(High-resolution, HR)内窥镜图像退化为符合真实域的低分辨率图像.同时,提出一种高频引导的残差密集超分辨网络,采用基于双频率信息交互的频率分离策略,并设计多层级融合机制,将提取的多级高频信息逐层嵌入残差密集模块的多层特征,以充分恢复内窥镜图像的高频细节和低频内容.在合成与真实数据集上的大量实验表明,我们的方法优于对比方法,具有更好的主客观质量评价.

基于深度学习的多帧遥感降质图像三维重建算法

为提升多帧遥感降质图像对比度以及图像质量,提出一种基于深度学习的多帧遥感降质图像三维重建算法。采用三角函数变换方法并结合高通滤波器,增强多帧遥感降质图像对比度;再以包含生成器和判别器的生成对抗网络为基础,在判别器中引入自注意力层,设计自注意力机制残差模块,生成自注意力生成对抗网络模型;最后将增强后的图像输入模型进行学习和训练,获取多帧遥感降质图像的全局特征后,实现多帧遥感降质图像三维重建。测试结果表明,所提算法具有较好的多帧遥感降质图像增强能力,能够提升图像对比度,并且渗透指数(PI)均在0.92以上,重构效果良好。

Research on the Application of Super Resolution Reconstruction Algorithm for Underwater Image

Underwater imaging is widely used in ocean,river and lake exploration,but it is affected by properties of water and the optics.In order to solve the lower-resolution underwater image formed by the influence of water and light,the image super-resolution reconstruction technique is applied to the underwater image processing.This paper addresses the problem of generating super-resolution underwater images by convolutional neural network framework technology.We research the degradation model of underwater images,and analyze the lower-resolution factors of underwater images in different situations,and compare different traditional super-resolution image reconstruction algorithms.We further show that the algorithm of super-resolution using deep convolution networks(SRCNN)which applied to super-resolution underwater images achieves good results.