ID 原文 译文
1033 保边滤波被广泛应用于图像的艺术化风格转换，但目前的方法大都只能产生类似卡通的绘画风格，对于具有明显笔画形状的绘画风格转换都无能为力。 Edge-preserving filters have been extensive used to image artistic stylization, but most of the existing filterscan only generate cartoon-like style and can do nothing about the painting styles with clear brush stroke shape.
1034 文章在 kuwahara 滤波的基础上提出了一种尺度可控的各向异性kuwahara 滤波，可以生成具有明显笔画形状的艺术效果。 Based on ku-wahara filter, the paper proposes a multi-scale anisotropic kuwahara filter to generate some painting styles with obvious brushstroke shape.
1035 为了更好的模拟画家的绘画技法，文章以图像显著性作为控制信息来实时调整滤波核大小，可以产生丰富的笔画尺度。 To simulate the painting techniques, saliency is used to control the size of the filter kernel and can produce strokes with rich geometrical scales.
1036 基于该算法，文章还给出了一个视频的艺术化风格渲染框架。该渲染框架可以保持良好的帧间一致性，自动的将视频转换为水粉画或油画效果。 Based on the multi-scale anisotropic kuwahara filter, we also propose an artistic render-ing framework for video stylization, which can keep frame-to-frame coherence well and automatically convert video into gouache or oil painting effect.
1037 同时利用图像的边缘切方向场图、边缘梯度和显著性控制凹凸映射，产生油画和水粉画笔触的厚重感和层次感，能很好的模拟“厚涂技法”的艺术风格。约束二维有限脉冲响应( Finite Impulse Response，FIR) 滤波器，现有设计算法计算复杂度高。 To produce brush strokes a thick and layered feel, bump mapping is used to produce the stroke touch effects. The edge tangent vector field map, edge gradient magnitude and image saliency are combined to control the bump mapping, which can simulate the“impasto technique”well in oil or gouache paintings.
1038 针对二维FIR 滤波器的约束最小二乘设计，本文应用交替方向乘子法( Alternating Direction Method of Multipliers，ADMM) ，研究其并行优化方法。 For constrained two-dimensional ( 2-D) finite impulse response ( FIR) filters, the computational complexi-ty of the existing design algorithms is very high. Based on the alternating direction method of multipliers ( ADMM) , the par-allel optimization of constrained least-squares ( CLS) 2-D FIR filters was studied.
1039 通过模型的最大分划，并采用一种松弛技术，提出一个具有高度并行结构的最大分划松弛 ADMM 算法，分析了算法的计算复杂度，讨论了算法的收敛性，并给出了算法的参数设置方法。 By maximally splitting the problem into univariate subproblems and utilizing a relaxation technique, a maximally split and relaxed ADMM with a highly parallelcomputing architecture was proposed. The computational complexity and convergence of the algorithm were analyzed, and apractical scheme for selection of the algorithm parameters was provided.
1040 实验表明，最大分划松弛 ADMM比非松弛的最大分划 ADMM 收敛快很多; Experimental results show that the proposed maxi-mally split and relaxed ADMM converges much faster than the maximally split unrelaxed ADMM.
1041 与现有算法相比，提高了计算效率。 Compared with existing algorithms, the computational efficiency of the proposed algorithm is improved.
1042 GPU 加速实验中获得的大加速比，表明了所提算法的高度并行性和可扩展性，在图像处理、计算机视觉、模式识别及机器学习等领域有广阔的应用前景。 The large acceleration ratios obtained byGPU demonstrate the high parallelism and scalability of the proposed algorithm, which is very valuable for applications of the algorithm in image processing, computer vision, pattern recognition and machine learning.