MPEG-7 based research on image shape features extraction and description method

摘要

In the researching field of content based image retrieval (CBIR), shape features based is the most difficult to deal with. The shape features are the most significant characteristics comparing with other lower basic image features and they are the hardest features to describe. In this paper, one method is presented to describe the features by adopting MPEG-7 standard recommend Fourier descriptor after the object shape features extracted by adaptive threshold segmentation. Choose the Fourier transformed border function to be shape descriptor. Use curvature scale function to represent shape features. Experiments show that image shape features extraction method used in this paper is accurate in edge location and will further facilitate the description of shape features. Meanwhile the use of MPEG-7 Fourier descriptors can express shape characteristics that contain abundant information with less data.