一种基于形态学滤波的视频对象空间分割新策略

视频对象提取是现代信息科学领域的一个研究热点。为了改进常见的视频对象时空分割方法,提取具有理想边缘的视频对象,本文提出了一种新颖的结合交替顺序重建滤波和自适应阈值判别的空间分水岭分割策略。该策略中结构元素逐渐变大的形态学开闭重建迭代使得分水岭分割所获区域数大大减少,有效地避免了复杂的区域融合。同时,迭代过程使得对象边缘像素的梯度与前景、背景中平坦区域内部像素的梯度更易区分,从而能够应用自适应阈值算法进一步消除由局部梯度极值造成的分割小区域。应用基于该策略的时空分割方法对标准测试序列进行分割实验,实验结果表明,该策略能够带来令人满意的空间分割结果,有助于提高视频对象时空分割方法的主客观性能。此外,该策略能够自动确定交替顺序重建滤波器中结构元素的大小以及非线性判别的阈值,算法的通用性和易用性较好。

A Novel Video Object Spatial Segmenting Strategy Based on Morphological Filtering

Video object extraction has been a most discussed topic in the field of modern information processing. In order to improve the performance of the common temporal-spatial segmenting algorithms and to obtain video objects with ideal boundaries, a novel video object spatial watershed segmenting strategy is proposed in this paper. This strategy combines the merits of alternating sequential filtering by reconstruction and adaptive threshold. The reduction of the number of the segmented regions without complicated area merging is guaranteed by the iteration of morphological filtering of opening and closing by reconstruction with gradually expanding structural elements. Besides, the iteration makes it easier to distinguish the gradient of the object boundary pixels and that of the pixels inside the foreground and background plateaus. This process facilitates the application of the adaptive threshold algorithms to remove the small regions of local gradient extreme. Several comparative temporal-spatial segmenting experiments have been implemented on some standard testing sequences. The experimental results indicate that the proposed strategy has achieved satisfactory spatial segmentation and enhanced the performance of the temporal-spatial segmenting algorithm for the subject and the object. Additionally, appropriate structural element size of the morphological filter and the non-linear threshold values can be automatically generated. This ensures fairly good versatility and usability of the algorithm.