摘要
一般的视觉显著性区域提取方法都不能一致地突出显著性区域和给出显著目标定义明确的边界.因此提出一种基于频率域的显著性区域提取方法.该方法使用高斯差分滤波器的组合来保持原图像更多的频率信息,在频率域里利用颜色和亮度信息来预测中间-周围对比度.该方法不仅输出完整分辨率的显著图,而且解决了上述两个问题.对真实数据实验结果表明,该方法分割的查准率和查全率都好于多尺度分析方法和Itti方法.
General visual salient region extraction methods can not consistently highlight salient region and give well-defined boundaries of salient objects.Therefore,a salient region extraction method based on the frequency domain is proposed.This method uses a combination of DoG filters to retain more frequency content from the original images,and uses color and brightness information to estimate the center-surround contrast in the frequency domain.This method not only output full-resolution salient map,but also solve the above two problems.This method outperforms on the ground-truth evaluation and on the segmentation task by achieving both higher precision and better recall than multiscale analysis method and Itti＇s method.
出处
《枣庄学院学报》
2012年第5期129-134,共6页
Journal of Zaozhuang University
基金
淮安软件和服务外包公共服务平台(项目编号:HAP201010)
关键词
显著图
频率域
多尺度分析
图像分割
查全率
saliency map
level set
spectral residual
image segmentation
recall
