Abstract | Keywords | Fulltext | Export | TOC

Low-Level feature such as color, texture, and shape features represent the visual content of an image. Feature Extraction process play a key role in Content Based Image Retrieval (CBIR), where automatically extracted the features from all images in the database and query image. In this paper, different type of feature extraction methods are explored to test their effectiveness in retrieving images including Color Moment (CM) and Color Histogram (CH) descriptors as color feature. Texture is represented by Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) descriptors and finally, Canny Edge Detection (CED) and Hu’s Seven Invariant Moments descriptor as shape descriptor. A new approach to choose the most appropriate descriptors to represent the image as uniquely and accurately using the average of success method and compare between the performances of each descriptor is presented. For query image several transformations process like rotation, cropping, etc., is applied to 100 original images collected from Iraqi National Museum of Modern Art collection to demonstrate experimentally the efficacy of the proposed approach and promising results are reported.

CBIR --- Feature extraction --- Low-Level features --- Feature selection --- Average of success method