【摘要】 首先提出了一种新的二值图像多尺度表示方法.该方法针对人的视觉特点,从图像内物体的形心出发,通过建立局部坐标系,将图像内各物体在各自坐标系下进行较为"紧密"的收缩变换.该方法有别于以往方法,能保持源图像的对称性,并且具有形状乃至拓扑的基本保持.其次,本文又提出了图像多尺度表示的优化方案,该方案可看成评判图像表示优劣的准则,也可作为图像多尺度表示的数学描述.更多还原

【Abstract】 A new method of multiscale representation of binary images is presented. According to the feature of human version, the algorithm creates local coordinate systems at the centroids of objects in the image in question, and then applies compact shrinking transformation in these systems. Different from any other algorithms, the approach can preserve symmetry of the original image and even keeps its shape or topology. Besides, an optimization scheme of multiscale representation of images is given. It can be regarded as a measurement of the performance of a pyramid construction method or as a mathematical definition for multiscale representation.更多还原