Image Retrieval

I. Introduction Digital images consist of pixels. Each pixel represents the color of a point in the image. Rectangular pixel arrays are called bitmaps or digital images. Advances in image procurement and preservation technology have resulted in a surprising development of a very large and detailed image database [1]. A large amount of image data is generated everyday, such as digital photos, medical images, satellite images, etc. [2]. Image mining can automatically extract meaningful information from more and more image data.

An image retrieval system is a computer system for viewing, retrieving and retrieving images from a large digital image database. Powerful natural, geographical and medical image retrieval using supervised classifiers focused on extracted features has been proposed. We implemented gray level co-occurrence matrix (GLCM), scale invariant feature technology (SIFT) and moment invariant features, extracted features from natural images, and performed GLCM and Gabor feature extraction on medical images.

One of the most effective algorithms for shape-based image retrieval is the Scale Invariant Feature Conversion (SIFT) algorithm originally developed by David Lowe of the University of British Columbia in 1999. Takes a single image as input and returns a set of features of the detected image. In the SIFT algorithm, the image filtering is based on a Gaussian function. After image filtering, SIFT uses a Gaussian difference (DoG) pyramid for spot (keypoint) detection. An image feature descriptor (called a keypoint descriptor) is a 128 element feature vector, formed by the magnitude and direction of the gradient computed for the area surrounding the identified keypoint. (Row, 2004)

Image annotation is the process of automatically assigning titles or keywords to digital images. Application to image search system for searching images in database was found. Machine learning methods and algorithms apply to automatic image annotation. Clustering and classification is the most common method used in image annotation. Reinforcement learning is a machine learning algorithm that learns how the agent behaves in the environment by interacting with the environment. In the field of machine learning, much research has been done recently. It is mainly used in games and robotics engineering. This algorithm approach is different from other machine learning algorithms, supervised learning and unsupervised learning.