Image fuzzy enhancement algorithm based on contourlet transform domain

In order to solve the problems of low contrast, global darkness and noise amplification in some hyperspectral and remote-sensing images, a new image fuzzy enhancement method based on contourlet transform (CT) domain is proposed. The algorithm includes the following four steps. Firstly, the cyclic translation method is used to suppress the pseudo-Gibbs phenomenon caused by the lack of translation invariance of the CT. Secondly, a nonlinear gain function is designed to enhance and suppress the high-frequency coefficients adaptively. Meanwhile, the multi-scale Retinex with mixed gray function is used to process the low-frequency sub-band coefficients. Then, the inverse translation and linear averaging and the inverse CT are performed on the enhanced coefficients, and finally the improved fuzzy contrast is used to enhance the texture and edge of the image globally. The experimental results show that the proposed method can make the image texture details more prominent, and enhance the overall visual effect of the images. Furthermore, the absolute mean difference and PSNR of images are also greatly improved .

     

基于人眼视觉特性的红外图像细节增强

由于红外成像的机理和探测器的限制,使得红外图像具有对比度低以及边缘模糊的特点。导致红外图像不适合被人眼及机器观察。传统的红外图像增强方法由于没有考虑人眼的因素,导致仍然不适合被人眼观察。本文结合分层思想提出一种基于人眼视觉的红外图像细节的增强的方法。该方法首先利用双边滤波器将图像分成包含图像低频概貌信息的基础层和包含图像高频细节信息的细节层。基础层的处理是运用人眼视觉特性的方法将其映射到显示器可显示范围内。细节层的处理是应用了自适应增益的方法对细节层进行增强。最后将两层的处理结果进行合并量化到8-bit范围内。采用大量的红外图像对算法进行测试,实验结果表明本算法能够改善红外图像对比度低和边缘模糊的缺点。     

De-hazing and enhancement method for underwater and low-light images

Because underwater and low-light images have different characteristics, there are few methods to jointly improve the visibility of these images. This paper proposes a de-hazing and enhancement method for underwater and low-light images to describe the two types of images uniformly. Multi-scale retinex color recovery (MSRCR) and guided filtering methods are used for de-hazing; the proposed method of white balance fusion global guided image filtering (G-GIF), effectively solve the problems of dim light, color distortion, and loss of edge details. Experiments show that compared with other methods, this method can effectively solve the image exposure, and at the same time, it can better protect and enhance the image’s color saturation and edge texture details, thus achieving a very good visual effect.

     

基于单幅图像的真实感人体动画合成

由于单幅图像缺失三维信息以及完整的纹理信息,基于单幅图像的真实感三维人体动画合成极具挑战性。针对单幅图像三维信息缺失问题,提出了一种基于SMPL参数模型的三维人体几何重建方法。该方法以单幅图像为输入,先根据输入图像人体轮廓信息变形标准的SMPL参数模型分别生成与目标轮廓一致的正反面的三维几何模型,然后利用基于B样条插值的网格拼接融合算法拼接正反面三维几何,最后为了恢复正确的手部几何,利用基于B样条插值的网格拼接融合算法,将重建后的模型上错误的手部几何用标准SMPL参数模型上正确的手部几何替换。同时,针对单幅图像中纹理缺失的问题,提出了一个称为FBN（front to back network）的对抗生成网络,用于恢复被遮挡的人体背面纹理。实验结果表明,该方法生成的具有完整纹理的人体几何能够由3D运动数据驱动运动,生成具有高度真实感的三维人体动画。     

The Retinex-based image dehazing using a particle swarm optimization method

To the best of our knowledge, currently the physical model based method is still an ill posed problem. Additionally, the image enhancement approaches also suffer from the texture preservation issue. Retinex-based approach is proved its effectiveness in image dehazing while the parameter should be turned properly. Therefore, in this paper, the particle swarm optimization (PSO) algorithm is firstly performed to optimize the parameter and the hazed image is converted into hue, saturation, intensity(HSI) for color compensation, In the other hand, the multi-scale local detail upgrading and the bilateral filtering approaches are designed to overcome the dehazing artefacts and edge preservation, which could further improve the overall visual effect of images. Experimental results on natural and synthetic images by using qualitative analysis and frequently used quantitative evaluation metrics illustrate the approving defogging effect of the proposed method. For instance, in a natural image road, our method achieves the higher e for 0.63, γ for 3.21 and H for 7.81, respectively and lower σ for 0.04. In a synthetic image poster, the higher PSNR for 18.17 and SSIM for 0.78 are also acquired compared to other explored approaches in this paper. Besides, the results performed on other underwater and aerial images in this study further demonstrates its defog effectiveness.

     

基于梯度曲面面积与稀疏约束的图像平滑方法

针对纹理图像在平滑过程中低对比度边缘易丢失和纹理细节抑制不彻底等问题,提出基于梯度曲面面积与稀疏约束的图像平滑方法。首先,将图像视作三维空间中的二维嵌入曲面,再在此基础上分析图像的几何特征并提出梯度曲面面积约束正则化项,以提高纹理抑制性能;其次,根据图像的统计特性,建立L₀梯度稀疏与自适应梯度曲面面积约束的混合正则化约束图像平滑模型;最后,采用交替方向乘子法对非凸非光滑的优化模型进行高效求解。通过纹理抑制、边缘检测、纹理增强和图像融合等方面的实验结果可知,所提出的图像平滑算法克服了L₀梯度最小化平滑方法易造成的阶梯效应和欠滤波等缺陷,能够在去除大量纹理信息的同时保持并锐化图像显著的边缘轮廓。     

颜色恢复和边缘保持的低照度图像超分辨率重建方法

针对常规图像超分辨率重建方法应用于低照度环境下的图像时存在纹理信息丢失、颜色偏移失真和重建性能退化等问题,提出了一种颜色恢复和边缘保持的低照度图像超分辨率重建方法。在锚定邻域回归（ANR）的图像超分辨率重建方法基础上引入颜色恢复和边缘保持的照度增强函数,从而提高图像内容和边缘纹理的显著性;选择最小加权二乘滤波作为中心环绕函数（WLS）以抑制高频特征退化;同时针对YCbCr颜色空间的Y通道分量采用边缘保持的照度增强函数计算其反射分量,进一步增强边缘纹理特征。实验结果表明,所提方法获得了更好的视觉效果,相比于其他方法,该方法峰值信噪比（PSNR）提高了63.15%,结构相似度（SSIM）分别提高了46.86%,感知质量（PI）提高了4.12%。     

融合小波变换与改进PCNN的图像增强算法研究

为了解决方向对纹理图像细节增强的限制问题,提出一种融合小波变换与改进脉冲耦合神经网络（PCNN）的图像增强算法。该算法首先对图像进行二维离散小波变换,提取图像的高频分量图。然后将图像像素的局部梯度值作为链接强度系数,在动态阈值函数中加入侧抑制信号来改进脉冲耦合神经网络;并用改进的脉冲耦合神经网络对高频分量图进行增强。最后使用中值滤波对小波重构后的图像进行非线性平滑,实现纹理图像细节的增强。实验结果表明,该算法能够有效地减少图像细节增强时方向的限制。增强后,纹理图像的细节更加丰富,整体对比度也有一定的提高。     

Monge mapping using hierarchical NURBS

Texture mapping is an efficient and effective tool in computer graphics and animation. While computationally very cost-effective, texture mapping may produce non-realistic appearances of shapes in 3D environment, especially when viewing closely. To improve the realism of 3D modeling, bump mapping technique is developed to add details with the 3D models on top of texture mapping. Bump mapping, however, offers only simple and visual enhancement. Displacement mapping technique can further improve the localized detail of geometry. In this paper, Monge mapping technique is developed for detail and local shape modification of NURBS represented geometry in a 3D environment. Based on multiresolution and refinement schemes, Hierarchical NURBS (H-NURBS) is first investigated to design a mechanism for the purpose of carrying localized geometric information. Monge mapping on H-NURBS patch can be easily performed via simple cut-and-paste operation. Parametric control of the local shapes is developed to facilitate easier and better 3D local modeling.     

An efficient proposed framework for infrared night vision imaging system

This research presents new three proposed approaches to enhancement the visibility of the Infrared (IR) night vision images. The first proposed approach depends on Hybrid Adaptive Gamma Correction (AGC) with Histogram Matching (HGCHM). The second proposed approach stands up Merging Gamma Correction with Contrast Limited Adaptive Histogram Equalization (MGCCLAHE). The HM uses a reference visual image for converting of night vision images into daytime images. The third approach mixes the benefits of the CLAHE with the undecimated Additive Wavelet Transform (AWT) Using Homomorphic processing (CSAWUH). The quality assessments for the suggested approaches are entropy, average gradient, contrast improvement factor, Sobel edge magnitude, spectral entropy, lightness order error and the similarity of edges. Simulation results clear that the third proposed approach gives superior results to the two proposed approaches from entropy, average gradient, contrast improvement factor, Sobel edge magnitude, spectral entropy and the computation time perspectives. On the other hand, the second proposed approach takes long computation time in the implementation with respect to the two proposed approaches. The second proposed approach gives better results to the first proposed approach entropy, average gradient, contrast improvement factor, Sobel edge magnitude, and spectral entropy perspectives. The first proposed approach gives better results to the two proposed approaches from lightness order error and the similarity of edges perspectives.

     

融合视觉感知和等周理论的数码迷彩纹理提取

现阶段数码迷彩纹理的提取方法缺乏主观视觉感知,针对其提取效果和质量欠佳的问题,提出一种融合视觉感知和等周理论的数码迷彩纹理提取方法。此方法先利用人的视觉感知特性构造边权值函数,将等周率作为选择分割阈值的准则选取最小等周率所对应的候选阈值作为最终分割阈值,完成数码迷彩纹理的提取。在一系列自然纹理图像上的实验结果表明,与现有的几种经典多阈值分割方法相比,此方法的分割效果更好。     

基于非下采样剪切波变换与引导滤波结合的遥感图像增强

针对遥感图像中对比度低、细节信息缺失和边缘梯度保持能力较弱等问题,提出了一种基于非下采样剪切波变换（NSST）与引导滤波相结合的遥感图像增强算法。首先,原始图像通过NSST被分解成低频子带和高频子带两部分。然后,对低频子带进行线性增强,提高整体对比度;采用自适应阈值法抑制高频子带的噪声,再对去噪后的高频子带进行引导滤波增强,提高图像的细节信息和边缘梯度保持能力。最后,对两部分子带进行NSST反变换,得到增强后的图像。实验结果表明,与直方图均衡、基于Contourlet变换和模糊理论的图像增强算法、基于非下采样Contourlet变换与反锐化掩膜结合的遥感图像增强算法以及基于非下采样Shearlet变换与参数化对数图像处理相结合的遥感图像增强算法相比,该算法的图像信息熵、峰值信噪比（PSNR）和结构相似性（SSIM）都有一定的提升,能明显地改善图像视觉效果,使得图像纹理更加清晰。     

Extra detail addition based on existing texture for animated news production

Animated news proposed by Next Media Animation becomes more and more popular because an animation is adapted to tell the story in a piece of news which may miss visual and audial circumstances. In order to fulfill the requirement of creating a 90-second animation within 2 h, artists must quickly set up required news elements by selecting existing 3D objects from their graphics asset database and adding distinguished details such as tattoos, scars, and textural patterns onto selected objects. Therefore, the detail addition process is necessary and must be easy to use, efficient, and robust without any modification to the production pipeline and without addition of extra rendering pass to the rendering pipeline. This work aims at stitching extra details onto existing textures with a cube-based interface using the well-designed texture coordinates. A texture cube of the detail texture is first created for artists to manipulate and adjust stitching properties. Then, the corresponding transformation between the original and detail textures is automatically computed to decompose the. detail texture into triangular patches for composition with the original texture. Finally, a complete detail-added object texture is created for shading. The designed algorithm has been integrated into the Next Media Animation pipeline to accelerate animation production and the results are satisfactory     

基于视觉感知的图像放大

为了有效保持图像纹理细节和边界结构,提出了一种基于视觉感知的图像放大算法。构造了一类[C2]连续的双变量有理函数插值模型。根据等值线生成原理,利用等值线方法将图像自适应地划分为边缘区域和平滑区域。基于图像的区域特征,平滑区域采用多项式模型插值,边缘区域采用有理模型插值。根据人眼对比敏感特性,对边缘区域进行参数优化,从而获得高质量插值图像。与其他经典的图像放大方法相比,该方法能有效保持图像的纹理细节和边缘结构,同时获得了较好的客观评价数据。     

小波变换与纹理合成相结合的图像修复

目的 为了克服传统的图像修复算法在结构和纹理边界的错误修复,利用小波变换域的系数特征,探讨了一种基于小波变换与纹理合成相结合的修复算法。方法 算法先利用小波变换将待修复图像分解成具有不同分辨率的低频子图和高频子图,然后根据不同子图各自的特征分别进行修复。对代表图像结构信息的低频子图,采用FMM(fast marching method)算法进行修复;对代表图像纹理信息的高频子图,根据各子图中小波系数的特征,利用纹理合成方法进行修复。结果 分层、分类修复方法对边缘破损具有良好的修复效果,其峰值信噪比相比于传统算法提高了1~2 dB。结论 与相关算法相比,本文算法的综合修复能力较好,可以有效修复具有较强边缘和丰富纹理的破损图像,尤其对破损自然图像的修复,修复后图像质量得到较大提升,修复效果更符合人眼视觉效应。     

基于稀疏分解的图像修复方法

提出将冗余离散小波变换(RDWT)和波原子变换(WAT)作为字典对应用于图像稀疏形态成分分解,获得图像的结构与纹理;然后针对结构和纹理所具有的不同形态学特征,对结构采用具有曲率驱动、边缘强化和平滑去噪性能的CDD模型修复,对纹理采用基于样例的Criminisi纹理合成方法修复;最后合成获得修复结果。实验结果表明,该修复方法能够获得强且光顺的边缘,纹理清晰完整,相比于传统方法具有更好的修复结果。     

级联离散小波多频带分解注意力图像去噪方法

针对图像去噪网络中下采样导致高频信息损失和细节保留能力差的问题,设计了一种级联离散小波多频带分解注意力图像去噪网络。其中多尺度级联离散小波变换结构将原始图像分解为多个尺度下的高低频子带来代替传统下采样,能减少高频信息损失。多频带特征增强模块使用不同尺度的卷积核并行处理高低频特征,在子网络每一级下重复使用两次,可增强全局和局部的关键特征信息。多频带分解注意力模块通过注意力评估纹理细节成分的重要性并加权不同频带的细节特征,有助于多频带特征增强模块更好地区分噪声和边缘细节。多频带选择特征融合模块融合多尺度多频带特征增强选择性特征,提高模型对于不同尺度噪声的去除能力。在SIDD和DND数据集上,所提方法的PSNR/SSIM指标分别达到了39.35 dB/0.918、39.72 dB/0.955。实验结果表明,所提方法的性能优于主流去噪方法,同时具有更清晰的纹理细节和边缘等视觉效果。     

基于模糊逻辑的指纹图像对比度增强算法

提出了一种新的基于模糊逻辑的方向纹理对比度增强的方法,对指纹图像进行增强.通过模糊边线判别器在对任一像素点属于边缘的程度进行模糊软判决的同时完成边缘检测,根据计算出的指纹方向图,结合模糊的边缘判新局面器的输出结果，来指导基于方向纹理的指纹图像的对比度增强.实验结果表明这种方法对于质量比较差的指纹图像能明显改善其质量,对于图像的边缘区域，则能够在保持边缘细节的同时滤除噪声。     

梯度优化的有理函数图像插值

目的 对图像纹理区域的细节保持一直以来是图像插值技术的一个难题,为此提出了一种梯度优化的有理函数图像插值算法。方法 首先,构造了一种新的含有可调参数的双变量有理插值函数,随着参数的不同取值,该函数具有不同的表达形式,它是多项式模型和有理模型的有机统一体;其次,根据图像的区域特征,利用等值线方法将图像自适应地划分为纹理区域和平滑区域,纹理区域采用有理模型插值,平滑区域采用多项式模型插值;最后,根据各向同性Sobel算子计算插值单元的图像梯度,确定纹理方向,不同纹理方向的插值单元用相应的权重对中心点进行优化。结果 从客观数据、主观效果、时间复杂度3个方面对重建图像进行评价,客观数据包括峰值信噪比（PSNR）和结构相似性（SSIM）,从实验结果可以看出,本文算法的PSNR平均提高了0.14~1.50 dB,SSIM平均提高了0.005~0.097。从主观效果来看,本文算法的重建图像的纹理细节更加丰富,边缘结构更加清晰,从时间复杂度来看,本文算法的平均运行时间是3.77 s,分别比DFDF（directional filtering and data fusion）、NEDI（new edge-directed interpolation）、RSAI（robust soft-decision adaptive interpolation）、Lee''s、NARM（nonlocal autoregressive model）算法快了3.28倍、5.26倍、53.28倍、43.53倍、418.54倍。特别地,对于Baboon、Barbara、Metal这类纹理细节丰富的图像,本文算法在峰值信噪比和结构相似性上较对比算法有突出优势,主观效果有明显提高。结论 基于构造的双变量有理插值模型,本文提出了一个梯度优化的有理函数图像插值算法,实验结果表明,该算法在图像纹理细节和边缘结构保持方面具有良好的视觉效果,有效提高了插值图像质量,且时间复杂度较低。     

Two low illuminance image enhancement algorithms based on grey level mapping

Two image enhancement contrast methods are proposed in this paper for low-intensity images. The first method (LEAM) is a new greyscale mapping function, and it can be significantly enhanced in the low grey range and compressed slowly in the high grey range, which is beneficial for retaining more image details; the second method (LEAAM) is based on the data characteristics of a histogram combined with the first mapping function, which adaptively sets the gamma value to correct the image. The experimental results show that compared with a traditional mapping function, LEAM is more effective at enriching image details and enhancing visual effects, and LEAAM, compared with a recent low-illumination image enhancement algorithm, achieves good performance for average gradient, information entropy and contrast index; additionally, the overall visual effect is the best compared with other methods.