Intelligent diagnosis of retinal vein occlusion based on color fundus photographs

AIM:To develop an artificial intelligence(AI)diagnosis model based on deep learning(DL)algorithm to diagnose different types of retinal vein occlusion(RVO)by recognizing color fundus photographs(CFPs).METHODS:Totally 914 CFPs of healthy people and patients with RVO were collected as experimental data sets,and used to train,verify and test the diagnostic model of RVO.All the images were divided into four categories[normal,central retinal vein occlusion(CRVO),branch retinal vein occlusion(BRVO),and macular retinal vein occlusion(MRVO)]by three fundus disease experts.Swin Transformer was used to build the RVO diagnosis model,and different types of RVO diagnosis experiments were conducted.The model’s performance was compared to that of the experts.RESULTS:The accuracy of the model in the diagnosis of normal,CRVO,BRVO,and MRVO reached 1.000,0.978,0.957,and 0.978;the specificity reached 1.000,0.986,0.982,and 0.976;the sensitivity reached 1.000,0.955,0.917,and 1.000;the F1-Sore reached 1.000,0.9550.943,and 0.887 respectively.In addition,the area under curve of normal,CRVO,BRVO,and MRVO diagnosed by the diagnostic model were 1.000,0.900,0.959 and 0.970,respectively.The diagnostic results were highly consistent with those of fundus disease experts,and the diagnostic performance was superior.CONCLUSION:The diagnostic model developed in this study can well diagnose different types of RVO,effectively relieve the work pressure of clinicians,and provide help for the follow-up clinical diagnosis and treatment of RVO patients.

Cross-CNN:基于CNN和Transformer混合模型的动画跨帧线稿着色算法

对长序列的动画线稿帧进行着色是计算机视觉中一项具有挑战性的任务.一方面,线稿中包含的信息较为稀疏,需要着色算法对缺失的信息进行推断;另一方面,连续帧之间的色彩需要保持一致,以确保整个视频的视觉质量.现有的着色算法多数只针对单张图片进行着色,这类算法只给出一个开放性的符合合理范围的色彩结果,无法适用于帧序列着色.另一些基于参考帧的着色算法,并没有将2帧之间的关系有机地联系起来,导致着色效果不够出色.在同一镜头序列中,同一对象的特征往往不会发生太大变化,因此,可以设计一个根据给定参考帧,即可给线稿自动着色的模型.为此,本文提出了基于CNN(Convolutional Neural Networks)和Transformer相结合的模型Cross-CNN,该模型能够从参考帧中寻找并匹配颜色,从而保证时间维度上的特征一致性.Cross-CNN模型参考帧和线稿帧在通道维度叠加,输入预训练的Resnet50网络提取局部融合特征,将融合特征图传给Transformer结构进行编码以提取全局特征.在Transformer结构中设计了交叉注意力机制更好地匹配远距离特征.最后使用带有跳层连接的卷积解码器完成着色图片输出.本文在数据集方面从8部电影中截取画面并经过严格筛选,最终制作了一个包含20000对二元组的数据集用于实验研究.Cross-CNN的SSIM(Structural SIMilarity)达到了0.932,高于SOTA算法0.014.本文算法代码链接:https://github.com/silenye/Cross-CNN.

基于Transformer网络特征融合的色纺织物颜色表征模型

针对色纺织物特有的呈色结构以及常用单一颜色测量工具导致的数据特征局限的问题,基于Transformer网络建立多源异构数据融合颜色表征模型。从分光计数据与图像数据中分别提取光谱特征与纹理特征,使用Transformer网络融合多源异构数据特征,充分利用其互补特性,有效完备地表征色纺织物颜色信息。结果表明,本文构建的颜色表征模型能够对在较大变化范围内的染色纤维质量配比差异和由于染色纤维表面不均匀分布的颜色变化进行有效表征;在分光光度仪6、10和25 mm测量孔径下,融合特征差异与纤维配比差异的相关系数均高于85%;相较于单一光谱特征和单一图像特征,本文方法的相关系数提升10%以上,具有理想的鲁棒性。

基于Transformer网络特征融合的色纺织物图像检索

针对单一特征难以准确描述色纺织物图像颜色和纹理的复杂性和各向异性,提出了基于Transformer网络特征融合的色纺织物图像检索算法。首先,利用卷积神经网络模型提取色纺织物图像的高级语义信息;然后,利用Transformer网络融合图像的三阶颜色矩特征,充分利用色纺织物图像的“高级语义信息”和“浅层图像信息”的互补性进行图像检索。采用了14种不同类型的色纺织物样本图像进行检索,该系统的平均Top-10查全率与准确率mAP值分别达到了98.35%和89.25%,相较于融合注意力机制的单一网络模型,检索Top-10的查全率和mAP值均有提升。

A color image encryption scheme based on a 2D coupled chaotic system and diagonal scrambling algorithm

A novel color image encryption scheme is developed to enhance the security of encryption without increasing the complexity. Firstly, the plain color image is decomposed into three grayscale plain images, which are converted into the frequency domain coefficient matrices(FDCM) with discrete cosine transform(DCT) operation. After that, a twodimensional(2D) coupled chaotic system is developed and used to generate one group of embedded matrices and another group of encryption matrices, respectively. The embedded matrices are integrated with the FDCM to fulfill the frequency domain encryption, and then the inverse DCT processing is implemented to recover the spatial domain signal. Eventually,under the function of the encryption matrices and the proposed diagonal scrambling algorithm, the final color ciphertext is obtained. The experimental results show that the proposed method can not only ensure efficient encryption but also satisfy various sizes of image encryption. Besides, it has better performance than other similar techniques in statistical feature analysis, such as key space, key sensitivity, anti-differential attack, information entropy, noise attack, etc.

基于Swin Transformer的疟疾细胞图像识别研究

为了协助医务人员更准确、更快速地诊断疟疾,提出一种基于Swin Transformer(SwinT)的疟疾细胞图像识别方案。方案采用伪彩色图像增强算法对血片图像进行预处理,以突出图像的颜色对比度,并引入SwinT模型作为主干网络,解决下采样固定和全局信息无法交互的问题,同时引入卷积层对图像进行线性变换,构建残差网络解决梯度消失和梯度爆炸问题。实验表明,与图像量化等其他图像增强方法相比,本文方法增强了疟疾细胞图像的色彩对比度,改进后方案的准确率达到99.7%,高于现有文献方法,可以对疟疾的辅助治疗带来更有价值的支持。

Assessing Gray Scale Grades for Color Change of Dyed Fabrics via Image Processing Technology

Gray scale grades for color change of dyed fabrics are assessed via image processing technology.Digital images of groups of specimens are obtained,cropped,and saved as JPEG format.Relationships between gray scale grades for color change and the corresponding color differences calculated via image processing technology are investigated,compared with those obtained from high accurate computer color matching system.Results show that the new method is acceptable with an accuracy of 92.0% when the grading errors are of not more than one grade.

False Color Fusion for Multi-band SAR Images Based on Contourlet Transform

为合成的孔雷达(SAR ) 图象基于 contourlet 转变的多乐队的一个假颜色图象熔化方法被介绍。它以下列方法工作：第一， contourlet 变换的灵活 multiresolution，高 directionality 和 anisotropy 被用于为 SAR 图象实现多尺度的分解。在 contourlet 变换领域，一条边信息测量规则被用来合并方向性的 subbands，并且一条平均规则被用来合并 lowpass subbands。然后，一个混合高增加过滤器算法被建议生产 red-green-blue (红绿蓝) 的隧道基于灰色的熔化图象建模的红、绿、蓝的颜色。最后，假颜色熔化了图象在 RGB 颜色空间被显示。这个方法把灰色的信息翻译成对人的视觉系统可得到的颜色信息并且提高光谱为 SAR 图象的分辨率。试验性的结果被证实建议方法的有效性。

Transforming the Snapdragon Aurone Biosynthetic Genes into Petunia Alters Coloration Patterns in Transgenic Flowers

Aurones belong to a small class of flavonoids that provide yellow color in some floricultural plants including snapdragon. To explore novel flower coloration, two full-length cDNAs encoding chalcone 4'-O-glucosyltransferase (designated as SRY4'CGT) and aureusidin synthase (designated as SRYAS1) in the aurone biosynthetic pathway were cloned from yellow flowers of snapdragon (Antirrhinum majus cv. Ribbon Yellow). Binary vectors were constructed and transformed separately into Petunia hybrida harboring blue flowers. Only a few flowers in 4 out of 9 transgenic SRY4'CGT plants showed variegated blue-white sectors;as time passed, amounts of variegated flowers and proportion of white sectors in the background blue color of the new-born flowers gradually increased, until finally, the petal color was completely white in all late-born flowers. In contrast, a few flowers in 3 out of 13 transgenic SRYAS1 plants showed variegated blue-white sectors;but, the amounts of variegated flowers did not increase over the whole flowering stage, and no complete white flowers were observed. RNA samples isolated from blue and white sectors of T1 transgenic SRY4'CGT plants were analyzed by reverse transcription-PCR, transgenic SRY4'CGT transcripts were detected in both sectors;however, transcripts of an upstream gene, chalcone synthase (CHS), were abundantly detected in the blue sectors but largely reduced in the white sectors, suggesting that the expression of CHS gene was suppressed in white sectors of transgenic plants. Furthermore, HPLC coupled with mass spectrometry demonstrated cyandin, malvidin and their derivatives were absent in white sectors, causing the white phenotype. Our findings may be attractive to molecular breeders.

A Watermarking Algorithm Based on Wavelet and Hadamard Transform for Color Image

Digital image watermarking is a useful solution to the problem of information security, copyright and network security. In this paper, we propose a watermarking algorithm for color image based HT and DWT. A binary image as watermark is embedded into green component or blue component of color image. The algorithm can satisfy the transparence and robustness of the watermarking system very well. The experiment based on this algorithm demonstrates that the watermarking is robust to the common signal processing techniques including JPEG compressing, adding noise, low pass filter, and mosaic.

An Analysis of Celie's Psychological Transformation in The Color Purple

Alice Walker is widely recognized as one of significant African-American writers and her most prominent work, The Color Purple, has won three awards altogether: the Pulitzer Prize, the America Book Award as well as the National Book Critics Circle Award nomination. It tells the story of how a submissive black girl named Celie undergoes the double oppression of racism and sexism at first, but transforms into an economically andmentally independent black woman eventually. This thesis will analyze Celie's psychological transformation from the perspective of Freud's personality theory and concludes that internal factors for Celie's psychological transformation areof great referential significance for all the women to explore their way to liberation.

Hybrid Color Texture Features Classification Through ANN for Melanoma

Melanoma is of the lethal and rare types of skin cancer.It is curable at an initial stage and the patient can survive easily.It is very difficult to screen all skin lesion patients due to costly treatment.Clinicians are requiring a correct method for the right treatment for dermoscopic clinical features such as lesion borders,pigment networks,and the color of melanoma.These challenges are required an automated system to classify the clinical features of melanoma and non-melanoma disease.The trained clinicians can overcome the issues such as low contrast,lesions varying in size,color,and the existence of several objects like hair,reflections,air bubbles,and oils on almost all images.Active contour is one of the suitable methods with some drawbacks for the segmentation of irre-gular shapes.An entropy and morphology-based automated mask selection is pro-posed for the active contour method.The proposed method can improve the overall segmentation along with the boundary of melanoma images.In this study,features have been extracted to perform the classification on different texture scales like Gray level co-occurrence matrix(GLCM)and Local binary pattern(LBP).When four different moments pull out in six different color spaces like HSV,Lin RGB,YIQ,YCbCr,XYZ,and CIE L*a*b then global information from different colors channels have been combined.Therefore,hybrid fused texture features;such as local,color feature as global,shape features,and Artificial neural network(ANN)as classifiers have been proposed for the categorization of the malignant and non-malignant.Experimentations had been carried out on datasets Dermis,DermQuest,and PH2.The results of our advanced method showed super-iority and contrast with the existing state-of-the-art techniques.

基于频率响应与图像特征提取的动车组变压器绕组状态诊断方法研究

动车组变压器是保障高速铁路稳定运行的核心设备,频率响应法是目前检测变压器绕组状态的有效方法。为提升车载变压器绕组状态诊断的准确性,结合暂态信号与频率响应法提出基于频率响应与图像特征提取的动车组变压器绕组状态诊断方法。搭建试验车载变压器绕组故障模拟平台,获取不同故障类型和故障位置的频响曲线,利用类Gram矩阵结合幅频和相频曲线信息,再利用密度分层法转换为伪彩色图,提取对应的灰度共生矩阵和灰度差分矩阵特征值,根据鹈鹕优化支持向量机方法对绕组故障进行诊断。试验结果表明:车载变压器绕组故障发生时,伪彩色图能够反映出故障信息,有利于图像分析和特征提取,采用基于频率响应与图像特征提取的动车组变压器绕组状态诊断方法能够识别车载变压器绕组的典型故障类型和位置。

分段均匀重分布和对数校正的水下图像增强

为了更有效地恢复水下图像的真实颜色、更好地改善图像的对比度和清晰度,提出了分段均匀重分布和对数校正的水下图像增强方法。用适中的图像强度分别对三通道图像进行强度补偿,对三通道分别进行分段均匀的灰度级重分布,以改善水下图像的颜色和提升图像的对比度;在Lab颜色空间中对亮度分量L进行自适应的对数斜率变换,以进一步改善水下图像的对比度和清晰度。实验结果显示,该方法在颜色恢复、对比度和清晰度的改善上优于部分现有方法,对水下图像增强后,获得更高的信息熵、UIQM和UCIQE值,因此,该方法具有更好的图像增强效果,对水下视觉具有更高的应用价值。

基于CRTNet的低照度图像增强方法

针对低照度环境下获取的图像与颜色失真问题,该文结合了通道注意力和空间注意力的机制,提出了一种基于颜色还原Transformer网络(color restoraration Transformer networks, CRTNet)的低照度图像增强方法。CRTNet由颜色注意力模块(color attention module, CAM)、颜色映射模块(color map module, CMM)和顺序增强结构组成。首先,CAM分为颜色通道注意力模块和颜色空间注意力模块两部分,利用Transformer的全局信息捕捉能力,颜色通道注意力模块关注有意义的颜色通道并赋予更高权重,生成通道注意力向量,颜色空间注意力模块使用三层卷积结构,关注高维空间中的空间细节信息,生成空间注意力权重图;其次,CMM通过线性拟合过程提取图像高维特征,对64D空间中的特征进行通道和空间两个维度的缩放和平移获得图像全局信息和细节信息,并与原始图像特征相结合,补充原始图像特征中颜色、亮度、对比度和细节等信息,实现颜色增强;最后,采用顺序增强结构,将CMM的输出作为输入重复进行3次CAM和CMM操作,以拟合更高阶的函数映射,实现低照度图像的有效增强。对公共数据集的实验和用户研究表明,所提方法在定量测量、细节与颜色复原方面取得了最好的结果。

基于Retinex的自适应水域图像去雾算法

针对利用Retinex理论对水域雾天图像去雾时出现去雾不彻底、色彩失真等问题,提出一种基于Retinex的自适应水域图像去雾算法,通过分析水域雾天图像特征,提出水域雾气浓度划分标准,实现自适应去雾。将水域雾天图像由RGB颜色空间转到HSI颜色空间,对亮度(Intensity,I)分量进行拉伸变换以调节亮度平衡,再用改进的单尺度Retinex(singal scale Retinex,SSR)算法对I分量进行增强处理,初步提高图像清晰度;将增强处理后的图像转换回RGB颜色空间进行自适应伽马变换增强,进一步提高图像的清晰度及对比度;利用HSI颜色空间进行图像饱和度补偿及其他操作实现对水域雾天图像的色彩恢复。实验结果表明,该算法能有效进行自适应水域雾天图像去雾,色彩恢复真实,可得到细节清晰的高质量去雾图像。

基于边缘检测的数字媒体深度交互式图像分割算法

针对数字媒体深度交互式图像受噪声干扰,导致其边缘检测效果较差,影响其分割精度的问题,提出基于边缘检测的数字媒体深度交互式图像分割算法。首先,利用小波变换方法对数字媒体中的图像进行去噪处理,提高图像分割精度;其次,使用高斯函数与低通滤波增强去噪后的图像,提高图像的清晰度,为图像分割提供便利;最后,依据自适应阈值算法对数字媒体的图像进行边缘检测,像素合集中存在上下阈值。根据其上下阈值的计算求出像素合集中的高低两个阈值,并对其实施边缘连接,实现数字媒体图像分割。实验结果表明,所提方法对分割数字媒体图像去噪效果好,分割精度和效率高。

Neural Network Method for Colorimetry Calibration of Video Cameras

To transfer the color data from a device (video camera) dependent color space into a device? independent color space, a multilayer feedforward network with the error backpropagation (BP) learning rule, was regarded as a nonlinear transformer realizing the mapping from the RGB color space to CIELAB color space. A variety of mapping accuracy were obtained with different network structures. BP neural networks can provide a satisfactory mapping accuracy in the field of color space transformation for video cameras.

Study on Tri-Stimulus Transformation

To decrease number of samples for the implementation of color space transformation, a method for modeling the chromatic characterization of video cameras was proposed. An additional transformation was required to predict output RGB values for an input color. This additional transformation was based on spectral reflectance relationship. The transformed color coordinates were taken as inputs of a multilayer neural network. Based on network outputs, the RGB values to be predicted were calculated. Experimental results were given to illustrate the performance of the method. Even though much less number of training samples are used, this method can also perform well on this color space transformation.

基于CF-YOLO的雾霾交通标志识别

针对现有交通标志检测模型在雾霾环境下出现漏检、错检以及参数较大等问题,设计一种基于YOLOv5s改进的CF-YOLO检测模型。为加强在雾霾环境中对交通标志的检测能力,提出一种基于颜色衰减先验的自适应伽马变换图像预处理算法;为增强对目标的定位能力及检测精度,将坐标注意力机制融合到网络中;为实现模型轻量化,引入FasterNetBlock构建网络。实验结果表明,改进算法在雾霾环境下交通标志检测相比原YOLOv5模型权重减少了2.3 MB,精度提高了8.5个百分点。