为了能更准确地检测出图像中的各种显著性目标，针对单一域的显著性目标检测方法容易造成有用信息遗失的缺点，提出一种将图像频域的全局信息与空域的局部信息结合获得更全面的显著性目标信息的新模型.通过融合两个域中的有用信息，将空域中的局部信息与频域中的全局信息进行信息的优势互补.此模型不但可以加强复杂背景下显著区域部分，而且可以有效抑制重复的非显著性部分.实验表明该模型方法相对于其它单一域的模型能较好的提取显著性目标区域，与单一的空频域模型比较，在准确率上相对于空域模型提高9．5％，相对于频域模型提高了6．3％.
In order to detect saliency goals more accurately,saliency detection method for single do-main is easy to cause the disadvantage of useful information loss.So a model is proposed by combining global information of the frequency and local information of the spatial to obtain more comprehensive saliency target information.The model is necessary.Through fusing the useful information of two do-mains,the final saliency detection combines the detail of the local information with the integrity of the global information.This model can not only strengthen saliency regions,but also can suppress non-significant parts in a complex background.Saliency detection which is obtained by the new model are clear and complete,and it can inhibit the background information to highlight the saliency regions. Experiments showed that the proposed model relative to the other single domain model can extract significant target area better.Compared with single domain model,the new model improved 9.5%relatively to spatial domain model,and improved 6.3% relatively to frequency domain model.
参考文献
| [1] | 杜杰;吴谨;朱磊.基于区域特征融合的RGBD显著目标检测[J].液晶与显示,2016(1):117-123. |
| [2] | 张颖颖;张帅;张萍;卢成.融合对比度和分布性的图像显著性区域检测[J].光学精密工程,2014(4):1012-1019. |
| [3] | Itti L.;Koch C..A model of saliency-based visual attention for rapid scene analysis[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,199811(11):1254-1259. |
| [4] | Li, Jian;Levine, Martin D.;An, Xiangjing;Xu, Xin;He, Hangen.Visual Saliency Based on Scale-Space Analysis in the Frequency Domain[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,20134(4):996-1010. |
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%