抽象的な

Artificial Neural Network Technique for CBIR Based On Query Image Feature Extraction

D.Santhosh, Tina Esther Trueman

In order to increase the Content-Based Image Retrieval (CBIR) system retrieval accuracy are obtain by applying effective clustering technique. It’s only focused on accuracy not time or space consumption. The main idea of this paper isto extract images feature such that shape, color and texture features for image retrieval process. HSV are calculated from the images. Then images are converted to grayscale and preprocessing is applied. Clustering of the images is done by Self-Organizing Map (SOM). From key image take out the color histogram then texture features. After that image shape will be determined by using edge detection method to an image. In other hand use shape filters to identify given shapes of an image. Latter the individuality of the global color histogram withlocal color histogram are compared first and find which is best among two characteristics. Then best histogram feature, texture features and shape feature are compared and explored for CBIR. By these works, the CBIR is created by using color, shape and texture fused features by means of constructing feature vectors weights. After that feature vector are cluster using Self-Organizing Map. The query image features are extracted and it compared with cluster. After that display the similar images which match with the cluster images. The proposed idea is to apply the neural network technique for clustering to increase the accurateness of image retrieval.