CBIR Features Extraction with GLCM, HC and HOG using SVM-RF Classifier
Anil Mishra · International Journal of Advanced Trends in Computer Science and Engineering · 2020
Content-Based Image Retrieval (CBIR)allows search and recovery of pictures that are similar to a known picture, by using attributes that represent the visual content of the pictures.Our proposed study aimed to model a method for the recovery of indexed pictures in databases from their visual content, without the need for textual annotations.Gray Level Co-Occurrence Matrix (GLCM), Harris Corner, and Histogram of Oriented Gradients (HOG) attributes are extracted from the publicly available databases (Corel and Caltech Datasets).A novel nonparametric method of texture combination is applied by means of Principal Components Analysis (PCA).Finally the classification is accomplished by using Support Vector Machine (SVM) and Random Forest Classifiers.The simulation outcomes show satisfactory performance accordant with accuracy, recall, precision and F-score.