QUANTIZED NEIGHBORHOOD LOCAL INTENSITY EXTREMA PATTERNS FOR IMAGE RETRIEVAL
L. Koteswara Rao, P Rohin, d P Sree Lakshmi · 2020
Technological advancements in the field of multimedia and digital electronics enable the users create millions of images every moment.Many researchers across the world developed algorithms to retrieve the images from a database.However, obtaining the better accuracy is still an open issue in image retrieval.Texture descriptive features like Local Binary Pattern (LBP) depend on the comparison of central pixel and its neighbors.However, the impact of the adjacent pixels is ignored in encoding the binary information.Our approach primarily relies on the fact that neighbors around a specified pixel hold considerable textural .A Horizontal-vertical-diagonal-anti diagonal structure is created from the image.Sign of difference in the intensity of pixels is used to form the patterns.We propose a novel texture descriptor, termed Quantized Neighborhood Local Intensity Extrema Pattern (QNLIEP).Retrieval efficiency of QNLIEP method is evaluated in the form of Precision as well as Recall.Standard image repositories such as ImageNet-25, Brodatz are used to evaluate the efficiency.