基于纹理特征的自适应插值

发布时间：2018-10-30 11:33

【摘要】：在传统有理插值函数的基础上构造出一种新的混合插值模型.该混合插值模型是有理函数与分形插值函数的有机整体,可由形状参数和尺度因子唯一确定.由于分形是刻画图像复杂度的有效工具,引入分形维数描述纹理的复杂程度.首先,提出一种基于局部分形维数的自适应阈值选取的方法,将整幅图像划分为纹理区域和非纹理区域.在纹理区域采用有理分形函数插值,在非纹理区域采用有理函数插值.尤其在有理分形插值模型中,提出一种基于分形维数的精确计算尺度因子的方法.最后,通过优化形状参数进一步提高插值图像质量.实验结果表明:提出的基于图像特征的混合插值模型与当前经典算法相比,尤其是在处理纹理图像方面,具有明显优势.
[Abstract]:Based on the traditional rational interpolation function, a new mixed interpolation model is constructed. The mixed interpolation model is an organic whole of rational function and fractal interpolation function, which can be uniquely determined by shape parameters and scale factors. Because fractal is an effective tool to describe image complexity, fractal dimension is introduced to describe the complexity of texture. Firstly, an adaptive threshold selection method based on local fractal dimension is proposed, which divides the whole image into texture region and non-texture region. Rational fractal function interpolation is used in texture region and rational function interpolation is used in non-texture region. In particular, in the rational fractal interpolation model, an accurate calculation method of scale factor based on fractal dimension is proposed. Finally, the interpolation image quality is further improved by optimizing shape parameters. Experimental results show that the proposed hybrid interpolation model based on image features has obvious advantages over the current classical algorithms, especially in the processing of texture images.
【作者单位】：山东财经大学计算机科学与技术学院;山东大学数学学院;山东大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(61373080,61672018,61402261,61373088,61272431,61332015) 山东省高等学校优势学科人才团队培育计划~~
【分类号】：TP391.41

【相似文献】