结构细化的神经风格迁移

风格迁移过程中风格元素均匀分布在整个图像中会使风格化图像细节模糊,现有的迁移方法主要关注迁移风格的多样性,忽略了风格化图像的内容结构和细节信息.因此,该文提出结构细化的神经风格迁移方法,通过增加边缘检测网络对内容图像的轮廓边缘进行提取实现风格化图像内容结构的细化,凸显内容图像中的主要目标;通过对转换网络中的常规卷积层的较大卷积核进行替换,在具有相同的感受野的条件下,使网络模型参数更少,提升了迁移速度;通过对转换网络中的常规卷积层添加自适应归一化层,利用自适应归一化在特征通道中检测特定样式笔触产生较高的非线性同时保留内容图像的空间结构特性来细化生成图像的结构.该方法能够细化风格化图像的整体结构,使得风格化图像连贯性更好,解决了风格纹理均匀分布使得风格化图像细节模糊的问题,提高了图像风格迁移的质量.     

目标边缘清晰化的图像风格迁移

针对图像风格迁移时出现前后景边界模糊造成风格化图像主要目标模糊的问题,提出了目标边缘清晰化的图像风格迁移算法.通过将用于提取内容图像轮廓的深度抠图神经网络与风格迁移网络合并,形成透明遮罩约束风格迁移过程,凸显风格化图像中主要目标的轮廓;通过对迁移网络中最大池化层进行替换,保留图像的背景信息,细化风格化图像的整体结构;通...     

基于深度学习的动漫风格迁移研究综述

图像动漫化技术的发展对我国动漫产业影响巨大.目前基于深度学习的动漫风格迁移研究是一项热门的研究方向,相关算法层出不穷.文章对动漫风格迁移领域现有的主流方法和代表性工作进行了归纳和讨论,分析了该领域所使用的主要深度神经网络模型,并按照动漫风格迁移方法所解决的不同实际问题,将其归纳为风景动漫迁移、人像动漫迁移和视频帧动漫迁...     

基于HSV颜色模型的图像风格迁移算法

风格迁移是通过使用图像处理方法将图像的内容与其他图像的色彩、纹理、轮廓等信息结合在一起的技术。它在保留图像原有内容的同时,还加入了其他艺术风格,最常使用的风格迁移技术就是深度学习技术。本文提出了一种基于HSV颜色模型的图像风格迁移算法,该算法是对经典的深度学习图像风格迁移算法进行改进。本文提出的算法利用HSV颜色模型在对颜色种类的表示的直观性和方便性的优势,在损失函数中加入内容图像与迁移图像的HSV颜色模型中的H因素的L2距离,并通过实验验证,本算法能够达到在风格迁移的基础上,保留原始内容图像颜色色调的目的。     

基于风格迁移的数据增强方法

针对深度学习模型在实际应用场景中预测性能下降的问题,提出了一种基于风格迁移的数据增强方法。首先,使用少量原始数据和少量实际应用场景下的未标注数据学习风格迁移模型。然后,对大量已标注的原始数据进行风格迁移,得到与实际数据风格相近的大量有标签数据。最后,基于此数据训练面向实际应用场景的深度学习模型。实验结果表明,所提出的方法能有效地提升模型在实际应用场景数据上的预测性能,且效果优于传统数据增强方法。     

基于快速风格迁移与Perlin噪声的虚拟地形生成方法

虚拟地形在游戏、动漫、VR和三维仿真等领域中有着广泛应用。针对传统Perlin噪声算法生成的地形细节较少、地貌失真等问题,本文提出了一种快速风格迁移(Fast Neural Style Transfer, FNST)与Perlin噪声相结合的生成方法。该方法根据地形特征划分区块并设定风格,使用VGG-16训练得到的变换网络进行风格迁移,并通过图像后处理保证区块边界平滑过渡。实验使用山地、平原、沙漠等不同地区的地形数据对模型进行差异训练,生成的虚拟地形与目标风格真实地形间峰值信噪比(PSNR)达到12.95 dB,结构相似性(SSIM)达到0.21,符合自然地形地貌,具有良好仿真效果。     

Image deraining with Adversarial Residual Refinement Network

Prior image deraining works mainly have two problems: (1) they do not generalize well to various datasets; (2) too much detail information is lost in the heavy rain area of the rain image. To overcome these two problems, we propose a new two-stage Adversarial Residual Refinement Network (ARRN) to deal with heavy rain images. Specifically, for the first problem, we first introduce a new implicit rain model to model a rain image as a composition of a background image and a residual image. Based on the proposed implicit model, we then propose the ARRN which consists of an image decomposition stage and an image refinement stage. For the second problem, a new attention Wasserstein Generative Adversarial Networks (WGAN) loss in the refinement stage is introduced to force the network to focus on refining heavily degraded areas. Comprehensive experiments demonstrate the effectiveness of the proposed approach.     

Deep Convolutional Neural Networks for pedestrian detection

Pedestrian detection is a popular research topic due to its paramount importance for a number of applications, especially in the fields of automotive, surveillance and robotics. Despite the significant improvements, pedestrian detection is still an open challenge that calls for more and more accurate algorithms. In the last few years, deep learning and in particular Convolutional Neural Networks emerged as the state of the art in terms of accuracy for a number of computer vision tasks such as image classification, object detection and segmentation, often outperforming the previous gold standards by a large margin. In this paper, we propose a pedestrian detection system based on deep learning, adapting a general-purpose convolutional network to the task at hand. By thoroughly analyzing and optimizing each step of the detection pipeline we propose an architecture that outperforms traditional methods, achieving a task accuracy close to that of state-of-the-art approaches, while requiring a low computational time. Finally, we tested the system on an NVIDIA Jetson TK1, a 192-core platform that is envisioned to be a forerunner computational brain of future self-driving cars.     

基于迁移学习的三子网图像去雾方法

目前大部分图像去雾算法只在一种或几种均匀雾图数据集中有较好的表现,对于不同风格或非均匀雾图数据集去雾效果较差,同时算法在实际应用中会因模型泛化能力差导致模型场景受限。针对上述情况,该文提出一种基于迁移学习的卷积神经网络(CNN)用于解决去雾算法中非均匀雾图处理效果不佳和模型泛化能力差等问题。首先,该文使用ImageNet预训练的模型参数作为迁移学习模型的初始参数,以加速模型训练收敛速度。其次,主干网络模型由3个子网组成：残差特征子网络、局部特征提取子网络和整体特征提取子网络。3子网结合以保证模型可从整体和局部两个方面进行特征提取,在现实雾场景(浓雾、非均匀雾)中获得较好的去雾效果。该文在模型训练效率、去雾质量和雾图场景选择灵活性3个方面进行了研究和改进,为衡量模型性能,模型选择在去雾难度较大的非均匀雾图数据集NTIRE2020和NTIRE2021上进行定量与定性实验。实验结果证明3子网模型在图像主观和客观评价指标两个方面都取得了较好的效果。该文模型改善了算法泛化性能差和小数据集难以进行模型训练的问题,可将该文成果广泛应用于小规模数据集和多变场景图像的去雾工作中。     

Neural Style Transfer for image within images and conditional GANs for destylization

In this paper, the feature representation of an image by CNN is used to hide the secret image into the cover image. The style of the cover image hides the content of the secret image and produce a stego image using Neural Style Transfer (NST) algorithm, which resembles the cover image and also contains the semantic content of secret image. The main technical contributions are to hide the content of the secret image in the in-between hidden layered style features of the cover image, which is the first of its kind in the present state-of-art-technique. Also, to recover the secret image from the stego image, destylization is done with the help of conditional generative adversarial networks (GANs) using Residual in Residual Dense Blocks (RRDBs). Further, stego images from different layer combinations of content and style features are obtained and evaluated. Evaluation is based on the visual similarity and quality loss between the cover-stego pair and the secret-reconstructed secret pair of images. From the experiments, it has been observed that the proposed algorithm has 43.95 dB Peak Signal-to-Noise Ratio (PSNR)), .995 Structural Similarity Index (SSIM), and .993 Visual Information Fidelity (VIF) for the ImageNet dataset. The proposed algorithm is found to be more robust against StegExpose than the traditional methods.     

A Novel Simple Visual Tracking Algorithm Based on Hashing and Deep Learning

Deep network has been proven efficient and robust to capture object features in some conditions.It still remains in the stage of classifying or detecting objects.In the field of visual tracking,deep network has not been applied widely.One of the reasons is that its time consuming made the strong method could not meet the speed need of visual tracking.A novel simple tracker is proposed to complete tracking task.A simple six-layer feed-forward backpropagation neural network is applied to capture object features.Nevertheless,this representation is not robust enough when illumination changes or drastic scale changes in dynamic condition.To improve the performance and not to increase much time spent,image perceptual hashing method is employed,which extracts low frequency information of object as its fingerprint to recognize the object from its structure.64-bit characters are calculated by it,and they are utilized to be the bias terms of the neutral network.This leads more significant improvement for performance of extracting sufficient object features.Then we take particle filter to complete the tracking process with the proposed representation.The experimental results demonstrate that the proposed algorithm is efficient and robust compared with the state-of-the-art tracking methods.     

基于矩的异分辨率图象边界检测

（１）本文引入仿射变换，提出了一种用矩算子以民分辨率图象直接进行边界检测的方法，利用同分辨率图象正方形窗口的模板行矩的计算，推导出了相应用的边界参数公式。（２）本文提出一种隐线笥插值矩计算方法，得到一组与正方形窗口对应的正方形模板，避免了费时的预插值过程，（３）对本文提出的方法进行了实验验证，并给出了实验结果。     

Photorealistic style transfer for video

This work proposes an efficient example-based photorealistic style transfer algorithm for video. Given a source unprocessed video and a reference image color graded by an artist, we describe an automatic algorithm to transfer the visual style of the reference onto the source footage. Our approach builds upon the color transfer methods based on the statistical properties of images. These methods are fast and of low-complexity, therefore suitable for real-time implementations. Our contribution is to adapt those methods to be used for unprocessed video from cinema cameras, and optionally to incorporate regions of interest previously selected by the user, affecting the final color transfer. The resulting videos are free from artifacts and provide an excellent approximation to the intended look, bringing savings in pre-production, shooting and post-production time. Results are free of spatio-temporal artifacts and the method outperforms diverse state-of-the-art methods according to observer preference experiments.     

基于深度卷积神经网络的气象雷达噪声图像语义分割方法

针对新一代多普勒气象雷达的散射回波图像受非降雨等噪声回波干扰导致精细化短时气象预报准确度降低的问题,该文提出一种基于深度卷积神经网络(DCNN)的气象雷达噪声图像语义分割方法。首先,设计一种深度卷积神经网络模型(DCNNM),利用MJDATA数据集的训练集数据进行训练,通过前向传播过程提取特征,将图像高维全局语义信息与局部特征细节融合;然后,利用训练误差值反向传播迭代更新网络参数,实现模型的收敛效果最优化;最后,通过该模型对气象雷达图像数据进行分割处理。实验结果表明,该文方法对气象雷达图像的去噪效果较好,与光流法、全卷积网络(FCN)等方法相比,该文方法对气象雷达图像中真实回波和噪声回波的识别准确率高,图像的像素精度较高。     

Edge detection in noisy images by neuro-fuzzy processing

A novel neuro-fuzzy (NF) operator for edge detection in digital images corrupted by impulse noise is presented. The proposed operator is constructed by combining a desired number of NF subdetectors with a postprocessor. Each NF subdetector in the structure evaluates a different pixel neighborhood relation. Hence, the number of NF subdetectors in the structure may be varied to obtain the desired edge detection performance. Internal parameters of the NF subdetectors are adaptively optimized by training by using simple artificial training images. The performance of the proposed edge detector is evaluated on different test images and compared with popular edge detectors from the literature. Simulation results indicate that the proposed NF operator outperforms competing edge detectors and offers superior performance in edge detection in digital images corrupted by impulse noise.