基于自适应局部阈值的交互式图像分割

摘要

Currently those rather popular interactive image segmentation techniques are sensitive to users’ inputting locations and the amount of inputs, moreover, the scribbles inputted by the user are tentative.Therefore, the user has to replace or add scribbles repeatedly in order to get satisfied result, which is quite complicated and trivial.In light of this issue, we propose a novel interactive image segmentation method.First, it generates a series of homogeneous regions and outlines their contours by using an image segmentation algorithm based on adaptive local threshold;then, the user roughly marks the region sets composing different objects with different colours; at last, the image segmentation can be completed by merging the region sets with same colour.Experimental results show that our algorithm meets user’ s need on more intuitional input and accurate result in interactive image segmentation, it also conveniently realises the image segmentation of multi-region and multi-object at the same time.%目前比较主流的交互式图像分割技术，对用户的输入位置和输入量比较敏感，而且用户的输入具有试探性，所以为了得到满意的分割结果，用户不得不多次调整自己的输入，比较繁琐。针对这种不足，提出一种新的交互式图像分割方法。首先采用基于自适应局部阈值的图像分割算法，生成一系列同质区域，并描绘出其轮廓；然后用户用不同颜色，粗略标记出组成不同目标的区域集；最后通过合并同色标记的区域集，完成图像分割。实验结果表明，该算法不仅满足用户对交互式图像分割输入更直观、结果更准确的需要，同时方便地实现了多区域多目标的图像分割。