Parallel structure-aware halftoning

Structure-aware halftoning technique is one of the state-of-the-art algorithms for generating structure-preserving bitonal images. However, the slow optimization process prohibits its real-time application. This is due to its high computational cost of similarity measurement and iterative refinement. Unfortunately, the structure-aware halftoning cannot be straightforwardly parallelized due to its data dependency nature. In this paper, we propose a parallel algorithm to boost the optimization of the structure-aware halftoning. Our main idea is to exploit the spatial independence during the evaluation of the objective function and temporal independence among the iterations. Specifically, we introduce a parallel Poisson-disk algorithm during the selection of pixel swaps, which guarantees the independency between parallel processes. Graphics processing unit (GPU) implementation of the technique leads to a significant speedup without sacrificing the quality. Our experiments demonstrate the effectiveness of the proposed parallel algorithm in generating structure-preserving bitonal images with much less time, especially for large images.     

Importance Driven Halftoning

Most halftoning techniques have been primarily concerned with achieving an accurate reproduction of local gray-scale intensities while avoiding the introduction of artifacts. A secondary concern in halftoning has been the preservation of edges in the halftoned image. In this paper, we will introduce a new halftoning technique that utilizes a bandpass pyramid to achieve an accurate reproduction of important attributes in the image. Ink is distributed through the bandpass pyramid primarily according to a user defined importance function. This technique has three main characteristics. First, our technique can produce results similar to many other halftoning techniques by allowing a generic importance function to be specified. If the chosen importance function is average intensity we obtain results similar to traditional halftoning. We also show how the importance function can be changed to highlight areas with high variance. Second, in addition to changing the importance function, the drawing primitives can also be changed. By using line segments instead of single pixels as drawing primitives we illustrate how edge enhancement can be achieved. Third, this technique allows the user to easily limit the number drawing primitives used. This is useful in limited resource rendering.
In addition to providing a tailorable halftoning technique our method can easily be adapted to produce two tone non-photorealistic (NPR) images. We illustrate this by showing how sketched effects can be achieved by aligning the drawing primitives according to different image attributes.     

Comprehensive Halftoning of 3D Scenes

The display of images on binary output hardware requires a halftoning step. Conventional halftoning algorithms approximate image values independently from the image content and often introduce artificial texture that obscures fine details. The objective of this research is to adapt a halftoning technique to 3D scene information and thus to enhance the display of computer generated 3D scenes. Our approach is based on the control of halftoning texture by the combination of ordered dithering and error diffusion techniques. We extend our previous work and enable a user to specify the shape, scale, direction, and contrast of the halftoning texture using an external buffer. We control texture shape by constructing a dither matrix from an arbitrary image or a procedural texture. Texture direction and scale are adapted to the external information by the mapping function. Texture contrast and the accuracy of tone reproduction are varied across the image using the error diffusion process. We halftone images of 3D scenes by using the geometry, position, and illumination information to control the halftoning texture. Thus, the texture provides visual cues and can be used to enhance the viewer's comprehension of the display.     

采用GPU的视觉定位系统加速实验

计算机视觉在实际应用中的主要瓶颈是系统的准确性和实时性,而实时性的提高又受限于计算机中处理器的处理速度。在一个视觉定位伺服实际系统中,要求有很高的实时性,尝试研究使用GPU建立实时的视觉系统的实际性。视觉定位系统软件分为图像粗定位(目标识别)、图像精定位、运动解算等几部分。当识别图像像素为640×480时,采用单独GPU加速的方式,比CPU加速了2.1444倍,采用CPU和GPU相结合的方式,比CPU加速了4.1548倍。     

Edge detection and edge-preserved compression for error-diffused images

In this paper, a new approach to edge detection and image compression of bilevel error-diffused images is proposed. The proposed approach is directly applied to the error-diffused images without any inverse halftoning technique. The main idea behind the proposed edge detection method is to compute the consistency value of each pixel of the error-diffused image, and the computed values are then clustered into two classes to obtain the desired edges. Here, the consistency value is a function of the mass center and geometric center of the window; more precisely, it represents the possibility that the area covered by the window can be uniform. As for compression, each error-diffused input image is compressed by dividing the image into non-overlapping blocks with size 8 × 8, then each highly-consistent block is encoded by its average illumination, and each lowly-consistent block is transmitted directly using the original bitmap. The threshold to distinguish these two kinds of blocks is automatically calculated based on the compression rate required by the users. The complexity of our compression technique is low, and this fast method can be used in real-time application.     

一种基于图像结构特征的实时点画生成算法

点画是一种经典的艺术技巧,其用点的形式模拟图像中色调的明暗变化,表现其 艺术特质。由图像自动生成点画是图像风格化和非真实感渲染的对象之一。为此,提出了一种 基于图像结构特征的实时点画生成算法。使用一个预计算的增量 Voronoi 序列快速生成点画所 需的采样点,通过输入图像和对应的结构特征图共同指导点画图中点的半径和色调,提高点画 图像在边缘及细节上的质量。实验结果表明,该算法生成的点画结果在视觉效果和数值误差 (SSIM)上均优于现有点画生成算法,并且在绘制速度上能够达到实时的性能,可应用于高帧率 的场景中。     

Neural network based method for image halftoning and inverse halftoning

A hybrid neural network based method for halftoning and inverse halftoning of digital images is presented. The halftone image is performed by single-layer perceptron neural network (SLPNN), and its corresponding continuous-tone image is reconstructed by radial-basis function neural network (RBFNN). The combined training procedure produces halftone images and the corresponding continuous tone images at the same time. The PSNR performance and visual image quality of these contone images achieved is comparable to the well-known inverse halftoning methods. The resultant halftone images compared with the error diffusion halftone are visually good, too. Furthermore, we apply different kinds of halftone images to a bi-level image compression method, called Block Arithmetic Coding for Image Compression (BACIC), which is better than the current facsimile methods.     

Interactive rendering of NURBS surfaces

NURBS (Non-uniform rational B-splines) surfaces are one of the most useful primitives employed for high quality modeling in CAD/CAM tools and graphics software. Since direct evaluation of NURBS surfaces on the GPU is a highly complex task, the usual approach for rendering NURBS is to perform the conversion into Bézier surfaces on the CPU, and then evaluate and tessellate them on the GPU. In this paper we present a new proposal for rendering NURBS surfaces directly on the GPU in order to achieve interactive and real-time rendering. Our proposal, Rendering Pipeline for NURBS Surfaces (RPNS), is based on a new primitive KSQuad that uses a regular and flexible processing of NURBS surfaces, while maintaining their main geometric properties to achieve real-time rendering. RPNS performs an efficient adaptive discretization to fine tune the density of primitives needed to avoid cracks and holes in the final image, applying an efficient non-recursive evaluation of the basis function on the GPU. An implementation of RPNS using current GPUs is presented, achieving real-time rendering rates of complex parametric models. Our experimental tests show a performance several orders of magnitude higher than traditional approximations based on NURBS to Bézier conversion.     

Digital halftoning by iterative isotropic error feedback

An error diffusion scheme for digital halftoning is proposed. The scheme is an iterative and isotropic error feedback process. Also, to compensate for the nonlinear tone reproduction by laser printers due to ink drop overlaps, we incorporate Pappas-Neuhoff's simplified printer model into the new error diffusion scheme. It appears that the proposed halftoning technique enhances the perceived tone continuity of halftone hard copies, and it alleviates the objectionable structured halftone textures of some existing methods. In addition, a mechanism is suggested to preview a halftone print on a gray-level CRT monitor. The user can interactively set the desired brightness of the halftone print and balance image contrast and tone continuity, which are two mutually exclusive subjective criteria for halftone reproduction.     

High-quality brightness enhancement functions for real-time reverse tone mapping

This paper presents an automatic technique for producing high-quality brightness-enhancement functions for real-time reverse tone mapping of images and videos. Our approach uses a bilateral filter to obtain smooth results while preserving sharp luminance discontinuities, and can be efficiently implemented on GPUs. We demonstrate the effectiveness of our approach by reverse tone mapping several images and videos. Experiments based on HDR visible difference predicator and on an image distortion metric indicate that the results produced by our method are less prone to visible artifacts than the ones obtained with the state-of-the-art technique for real-time automatic computation of brightness enhancement functions.     

图像逆半调及其质量评价综述

数字半调是数字印刷乃至现代印刷的核心技术之一。半调可使连续色调图像转换为二值图像。如需对半调图像作分类识别、图像压缩、图像增强等处理,就应先让半调图像重建为连续色调图像,即进行逆半调处理。首先介绍了半调图像的不同发展阶段;然后对已有的逆半调方法根据不同的图像恢复依据进行了分类介绍,并对重建的逆半调图像的质量进行了客观评价;最后总结了已有逆半调方法的不足,并指出了未来可改进的关键方向。     

Fast Gabor texture feature extraction with separable filters using GPU

Gabor wavelet transform is one of the most effective texture feature extraction techniques and has resulted in many successful practical applications. However, real-time applications cannot benefit from this technique because of the high computational cost arising from the large number of small-sized convolutions which require over 10 min to process an image of 256 × 256 pixels on a dual core CPU. As the computation in Gabor filtering is parallelizable, it is possible and beneficial to accelerate the feature extraction process using GPU. Conventionally, this can be achieved simply by accelerating the 2D convolution directly, or by expediting the CPU-efficient FFT-based 2D convolution. Indeed, the latter approach, when implemented with small-sized Gabor filters, cannot fully exploit the parallel computation power of GPU due to the architecture of graphics hardware. This paper proposes a novel approach tailored for GPU acceleration of the texture feature extraction algorithm by using separable 1D Gabor filters to approximate the non-separable Gabor filter kernels. Experimental results show that the approach improves the timing performance significantly with minimal error introduced. The method is specifically designed and optimized for computing unified device architecture and is able to achieve a speed of 16 fps on modest graphics hardware for an image of 256² pixels and a filter kernel of 32² pixels. It is potentially applicable for real-time applications in areas such as motion tracking and medical image analysis.     

Texture-based Dither Matrices

Continuous tone images must be halftoned to be displayed on binary output devices such as printers. Halftoning algorithms at low resolutions of the output hardware introduce textures into the resulting display. In this work we control halftoning texture by generating a threshold matrix from an image-based texture. We demonstrate that processing textures by the adaptive histogram equalization algorithm approximates pixel distribution properties of traditional dither screens. Ordered dithering with the resulting threshold matrix enables us to define texture in the halftoned image. We control the appearance of this texture by a combination of the ordered dither algorithm with an error diffusion process. We present applications of texture-based dither screens to both photorealistic and artistic rendering. In the case of photorealistic tone reproduction our technique preserves textures and edges of the original image. The ability to define an arbitrary texture enables us to introduce a variety of artistic effects, including embossing of images with textures and text, and approximation of the appearance of of conventional illustration media. We evaluate the resulting halftoning using multi-scale edge distortion measures. Our quantitative evaluation closely corresponds to the visual observations.     

Data-parallel agent-based microscopic road network simulation using graphics processing units

Road network microsimulation is computationally expensive, and existing state of the art commercial tools use task parallelism and coarse-grained data-parallelism for multi-core processors to achieve improved levels of performance. An alternative is to use Graphics Processing Units (GPUs) and fine-grained data parallelism. This paper describes a GPU accelerated agent based microsimulation model of a road network transport system. The performance for a procedurally generated grid network is evaluated against that of an equivalent multi-core CPU simulation. In order to utilise GPU architectures effectively the paper describes an approach for graph traversal of neighbouring information which is vital to providing high levels of computational performance. The graph traversal approach has been integrated within a GPU agent based simulation framework as a generalised message traversal technique for graph-based communication. Speed-ups of up to 43 × are demonstrated with increased performance scaling behaviour. Simulation of over half a million vehicles and nearly two million detectors at a rate of 25 × faster than real-time is obtained on a single GPU.     

Bit‐depth extension: Overcoming LCD‐driver limitations by using models of the equivalent input noise of the visual system

Abstract— Continuous tone, or “contone,” imagery usually has 24 bits/pixel as a minimum, with 8 bits each for the three primaries in typical displays. However, lower‐cost displays constrain this number because of various system limitations. Conversely, higher‐quality displays seek to achieve 9–10 bits/pixel/color, though there may be system bottlenecks limited to 8. The two main artifacts from reduced bit‐depth are contouring and loss of amplitude detail; these can be prevented by dithering the image prior to these bit‐depth losses. Our technique builds on Roberts's noise‐modulation idea and the subsequently influenced work in halftoning for hardcopy and dithering for displays. However, most halftoning/dithering work was primarily directed to displays at the lower end of bits/pixel (e.g., 1 bit as in halftoning) and higher ppi. We approach the problem from the higher end of bits/pixel/color, for example, 6–8, and lower spatial resolution (<100 ppi), which changes the game substantially from halftoning experience. Instead of spatial dither, it is better to use an amplitude dither. In addition, dynamic displays allow for the use of a temporal dithering component. This paper will report on techniques and observations made in achieving contone quality on ～100‐or‐less‐ppi LCDs starting from 4‐ to 8‐bit driver limits, and resulting with no visible dither patterns, noise, contours, or loss of amplitude detail at viewing distances as close as the near focus limit (～120 mm).     

GPU上的水彩画风格实时渲染及动画绘制

论文提出了一种基于GPU 的对三维场景进行实时水彩画效果渲染的方法。 该方法的大部分过程使用图像空间的技术实现。算法将画面分为细节层、环境层、笔触层分 别渲染,再进行合成。在过程中使用环境遮挡、shadow mapping 等技术进行快速的阴影计算, 并使用图像滤镜的方法模拟水彩的多种主要特征。由于该方法以图像空间的技术为主,因此 可以利用GPU 并行处理的特点对计算过程进行加速,进而达到实时的渲染速度。最后建立 动画脚本分析系统,进行实时动画渲染,表明该方法在计算机动画、游戏等数字娱乐产业领 域有较大的应用潜力。     

Scale-dependent and example-based grayscale stippling

We present an example-based approach to synthesizing stipple illustrations for static 2D images that produces scale-dependent results appropriate for an intended spatial output size and resolution. We show how treating stippling as a grayscale process allows us to both produce on-screen output and to achieve stipple merging at medium tonal ranges. At the same time we can also produce images with high spatial and low color resolution for print reproduction. In addition, we discuss how to incorporate high-level illustration considerations into the stippling process based on discussions with and observations of a stipple artist. Also, certain features such as edges can be extracted and used to control the placement of dots to improve the result. The implementation of the technique is based on a fast method for distributing dots using halftoning and can be used to create stipple images interactively. We describe both a GPU implementation of the basic algorithm that creates stipple images in real-time for large images and an extended CPU method that allows a finer control of the output at interactive rates.     

A character art generator using the local exhaustive search,with GPU acceleration

An ASCII art is a matrix of ASCII code characters that reproduces an original grey-scale image. A JIS art is an ASCII art that uses JIS Kanji code characters instead of ASCII code characters. They are commonly used to represent pseudo grey-scale images in text -based messages. Since automatic generation of high quality ASCII/JIS art images is very hard, they are usually produced by hand. The main contribution of this paper is to propose a new technique to generate an ASCII/JIS art that reproduces the original tone and the details of an input grey-scale image. Our new technique is inspired by the local exhaustive search (LES) to optimise binary images for printing based on the characteristic of the human visual system. Although it can generate high quality ASCII/JIS art images, a lot of computing time is necessary for the LES. Hence, we have implemented our new technique in a graphics processing unit (GPU) to accelerate the computation. The experimental results show that the GPU implementation can achieve a speedup factor up to 89.56 over the conventional CPU implementation.     

基于B样条小波的彩色印刷扫描图像逆半调

彩色印刷是一种利用黄、品红、青、黑四色油墨叠印产生彩色图像的技术,逆半调是将半色调图像转换成连续调图像的过程,主要应用在电子出版和印刷图像电子化方面。选取LAB颜色空间作为逆半调处理的颜色空间,选择三阶B样条小波作为小波基,遵循小波分解、小波系数调整、小波重构的小波变换算法,对彩色印刷扫描图像进行逆半调处理,得到逆半调图像。实验结果表明,该算法避免了破坏彩色图像红、绿、蓝颜色通道的颜色相关性,对彩色印刷扫描图像的重建和印刷加网图像的恢复具有一定的理论意义和应用价值。     

利用可编程GPU实现大规模地形场景的高性能漫游*

对已有算法进行了综述,并针对数据动态调度、自适应网格模型的生成以及数据的组织与数据裁剪等方面进行了研究并提出改进方法,设计了一种基于GPU编程实现的大规模地形场景的实时绘制与漫游算法。利用GPU端完成地形网格更新、地形块的自动选取、高度图和纹理图采样等大部分计算工作,大大减轻了CPU端的计算负载。实验表明,该算法实现简单,内存开销较少,有效提高了地形绘制的效率,适于大规模地形场景的实时高效漫游。