Performance analysis of shape, color and texture features on tracking information face based on CBIR

Abstract

The image has the features of shape, color and texture that are vary. Each feature has a different performance in supporting the accuracy of information retrieval using a process approach to CBIR (Content-Based Image Retrieval). On the image with different objects different performance will be generated on each feature. For example, that the performance features of the form of the more dominant compared features color and texture on the image with the face, while the object on the image with the object of interest feature is more dominant than the features of texture and shape. In this research was conducted on the analysis of the performance features of the shape, color and texture in supporting the accuracy of a search using the approach of CBIR (Content Based Image Retrieval). The method used are invariant moment, color moment and GLCM (Grey Level Co-occurrence Matrix). The results showed that the best search accuracy is 95%, where the features of shape has a performance by 50%, 30% color feature and texture feature by 20% with 600test image with object database face.