一种高效的基于局部特征的显著区域检测算法

图像特征提取是图像/视频检索问题的关键。研究人员发现,假如对整张图像或视频帧做特征提取,会产生大量的特征信息,从而影响图像/视频帧的匹配时间。由于人们通常只对图像中部分关键的区域感兴趣,因此在对图像做局部特征提取时可以先对图像进行显著区域检测,排除非关键区域上的特征信息,从而减低匹配所需的特征点数,提高匹配的处理速度。但是由于传统的显著区域检测算法计算复杂,会对图像特征提取过程产生额外的时间开销。提出一种快速的显著区域检测算法,根据局部特征提取算法特征检测阶段得到的特征点分布,利用最大子矩阵和算法,在损失较小的准确度的情况下快速检测出图像的显著区。

AN EFFICIENT LOCAL FEATURE-BASED SALIENT REGION DETECTION ALGORITHM

Image features extraction is the key for image/video retrieval issue.However,the researchers find that to extract features from a whole image or video frame will generate huge amount of feature information,which affects the matching time of image/video frame.As people are only interesting in part of the key region of the image,therefore,detecting salient region before the local features extraction on image is a good choice for excluding the feature information on non-key regions and to improve processing speed of matching by reduce the feature points the matching needed.Since traditional salient region detection algorithms are too complicated and result in extra time overhead in process of image features extraction,this paper introduces a fast salient-region detection algorithm.It extracts according to local features the feature points’ distribution derived in features detection stage of algorithm,and uses max sum of sub-matrix algorithm to fast detect the salient region of image in condition of less precision lost.