Research study of neural networks for image categorization and retrieval

In many areas of commerce, government, academia, and hospitals, large collections of digital images are being created. Many of these collections are the product of digitizing existing collections of analogue photographs, diagrams, drawings, paintings, and prints. Usually, the only way of searching these collections was by keyword indexing, or simply by browsing. Digital images databases however, open the way to content-based searching. In this paper we survey some technical aspects of current content-based image retrieval systems based on several neural network architectures. Firstly we discuss the image retrieval system based on neural network. The advantage of using the neural network is that the amount of semantic gap can be reduced .The methodology discussed below is designed for a specific class of objects, which can be broken down into sub objects in such a way that the main object can be classified by shape, color distribution and texture of the sub objects and the spatial relations between the sub-objects in a 2-dimensional image. We also assume that translation, scaling and 2D rotation do not change the class of the object, but we do not consider 3D transformation.Therefore, photos of the same 3D object from different positions for example are considered to be objects belonging to different class. The next approach which is discussed is based on Multiple Instance Learning approach and a neural network (a multiple-layer perceptron to be specific), which is trained to categorize images using back propagation.