基于图像变形技术的平行坐标可视化

摘要

For the drawbacks such as not suitable for processing high dimension data exist in conventional parallel coordi-nates, an improved method based on image morphing is proposed. Firstly, morph function is established between control-ling points and object points within study dataset. Then, the test datasets can be mapped on the parallel coordinates by using of the parameters from morph function. The better results of visualization and classification can be supported by the experiments.%针对传统的平行坐标可视化方法不适于处理大量高维数据等不足，将基于控制点的图像变形技术引入到可视化中，首先利用学习集中的数据建立控制点到目标点的变形函数，求出方程的参数，再将测试集中的数据通过变形函数映射到平行坐标系中，实验结果表明，通过图像变形原理能较好地实现高维数据的分类和可视化效果，优于传统的平行坐标方法。