Multi-tier content-based microscopic image retrieval for multi-image queries

Image Retrieval is a system of searching, surfing, and retrieving the images from an image dataset. There are two types of different image retrieval techniques namely text based image retrieval and content based image retrieval (CBIR) techniques. Text-Based image retrieval used to manage and retrieve images with traditional database techniques. Content-based image retrieval uses the color feature, shape feature, texture feature, and spatial layout to characterize and manifest the image. This CBIR system is greatly used application in medical filed to handle large set of diseased image database. In this paper, an efficient CBIR system and low-level feature extraction technique is used. This proposed system that gives an efficient retrieving system of microscopic images by multitier approach of CBIR system. Color features and texture features are the features extracted in low-level feature extraction. Normalization is applied over extracted features and for further content-based image retrieval (CBIR) process store the normalized feature. The first tier is classification of disease type with Adaptive neuro fuzzy inference system and the second tier is slide-level image retrieval that deals efficient microscopic image retrieval with k-nearest neighbor (KNN) to find grades of the disease images. The microscopic images taken for this paper are neuroblastoma (NB) and follicular lymphoma (FL) with their subtypes. When using ANFIS instead SVM classifier, the classification of disease images brings more accurateness for retrieval process.