基于直方图匹配的艺术Mosaic图像并行拼贴

Parallel Collage of Artistic Mosaic Images Based on Histogram Matching

为获得高质量的艺术Mosaic图像拼贴视觉效果,提出基于直方图匹配的并行拼贴算法。根据直方图距离挑选原图像分割块中最佳匹配子图像,采用直方图匹配法对艺术Mosaic图像进行颜色校正。拼贴算法中,原图像分割、直方图比较、颜色校正及子图像填充过程均使用基于CUDA(统一计算设备架构)的GPU(图形处理器)并行计算,通过实验分析和用户调查验证本文算法的有效性。结果表明:与传统基于颜色均值的颜色校正法相比,本文算法拼贴结果更接近原图,整体视觉效果更协调;基于CUDA的GPU并行计算提高了算法效率,从而实现交互式艺术Mosaic图像递归式拼贴。

In order to obtain high quality artistic Mosaic collage visual effect,a parallel collage algorithm based on histogram matching was proposed.According to the histogram distance,the best matching sub-image of the original image segmentation block was selected,then the histogram matching method was used for color correction.In the collage algorithm,segmentation of original images,comparison of histograms,color correction and image filling were all parallelly performed with GPU(graphic processing unit)based on compute unified device architecture(CUDA).The effectiveness of the proposed algorithm was verified by experimental analysis and user survey.The results show that compared with the traditional color correction method based on color mean,the collage result of this algorithm is closer to the original image,and the overall visual effect is more coordinated.GPU parallel computing based on CUDA improves the efficiency of the algorithm,thus realizing the recursive collage of interactive art Mosaic images.