【摘要】 在现有的基于块的纹理合成算法中,针对块尺寸需要人工选择从而导致纹理合成质量不确定的问题,提出一种纹理合成中的块尺寸自动选择算法。在纹理样本上按扫描线顺序滑动子块直到遍历所有位置,对子块与纹理样本的直方图进行归一化与均值滤波预处理,然后计算二者直方图的交;在不同子块位置的上述计算结果中,取其最大值作为子块与样本的颜色相似度。针对颜色相似度与块尺寸的近似单调递增关系,采用二分法计算相似度阈值点所对应的横坐标,将其作为纹理合成的块尺寸。多种类型纹理的实验结果表明,该方法自动选择的块尺寸与最佳经验取值范围相吻合。所提方法不仅适用于结构性纹理的合成,而且适用于随机性纹理的合成,能够获得理想的合成结果。更多还原

【Abstract】 In previous patch-based texture synthesis algorithm, the size of patch requires artificial choice which leads to an uncertainty quality of texture synthesis, so an automatic selection algorithm of patch size for texture synthesis was presented. When the patch was slid on the examplar in scan-line order until getting through all the plane, the histograms of the patch and the examplar were both pretreated by normalization and mean filtering, then the intersection of the two histograms was calculated. Among all the calculation results for different position, the maximum value was selected as the measurement of the color similarity between patches and examplar. Due to approximate monotone increasing relation between color similarity and patch’s size, bisection method was adopted to calculate the abscissa for the color similarity threshold point, and the abscissa was used as the patch’s size for texture synthesis. The experimental results of various types of textures show that the size of patch automatically selected by this method coincides with the range of best empirical value. Compared with the other automatic selection methods of patch’s size, this method not only applies to structured texture synthesis, but also applies to stochastic texture synthesis, and obtains suitable results of texture synthesis.更多还原