International Journal of Engineering, Science and Mathematics

Texture plays a significant role in most of the Content Based Image Retrieval systems (CBIR) when compared to other low level visual descriptors. Since there are plenty of methods available for extracting the texture, among which Local Binary Pattern (LBP) is considered the state of the art method for effectively describing the texture. It is not only used for describing the texture apart from that it has been applied to many applications in image and signal processing such as texture classification, face and facial expression recognition, object tracking and leaf image classification etc. As this considers spatial features only, a similar another method has been existed called local phase quantization which uses the frequency features of the texture by considering neighborhood of a pixel. The Local Binary Patterns (LBPs) are first proposed for representing the texture by encoding the pixel wise information. In this method, given a center pixel in the (3 × 3) window, LBP value is computed by comparing its neighborhoods grayscale value with center pixel. Then the neighboring pixels are assigned with a binary label, which can be either 0 or 1 depending on whether the center pixel has higher intensity value than the neighboring pixel. These binary values multiplied by specific weights and summed up. But they gives variant values when an image is rotated etc. hence an improved LBP has been used which gives same texture properties even on image rotation. After that GLCM metrices has been evaluated from both spatial and frequency descriptors. For retrieving the images by inputting a query image, For different types of similarity metrics has been used in which all four measures gives almost same accuracy rate in retrieving the images but Canberra distance gives highest accuracy on all types of images.