基于多特征图像显著目标的自动识别

现有的许多显著目标检测算法,大都依赖像素间的相互关系,而缺乏对焦点对象特征的理解.本文提出了一种无监督的显著目标自动识别算法.首先,应用具有仿生学特质的多尺度Gabor模型检测图像中的初级特征,构成显著目标的基本元素特征.在此基础上,结合显著目标的自身特性定义图像的局部特征,进一步确定感兴趣区域的位置.最后,在模糊阈值算法的基础上提出一种新的提取策略,并将其应用到由初级特征和局部特征构成的显著图上,从而准确地确定显著目标的位置.将此方法应用于具有不同特点的图像进行仿真实验,得到了较好的结果,证明该算法是切实可行的.

AUTOMATIC IDENTIFICATION OF SALIENT OBJECT BASED ON MULTI-FEATURES IMAGES

Many existing detecting algorithms of salient object mostly depend on the relationship between the pixels and lack the comprehension of salient object features.An unsupervised algorithm for automatically extracting the interesting object was presented.Firstly,the junior feature was extracted as basic elements of salient object by using a multiscale Gabor model with bionics characteristics.Then by combining with the understanding knowledge of the salient object,the local feature of image was defined and the ...