<title>Image indexing and retrieval techniques: past, present, and next</title>

Abstract

ABSTRACT This paper investigates image retrieval and indexing schemes in pixel domain, and points to the future work on image retrievalschemes. Image retrieval schemes generate indices for images in pixel or compressed domain based on their features in thecorresponding domain. These indices are used to retrieve images from a database. The features in pixel domain are extractedfrom the color, shape or texture characteristics ofimages. The application ofthese three methods depends on the characteristic ofthe image database, and the query image. In the near future the image databases contains the compressed version of images,therefore there will be a high demand for the image retrieval techniques in compressed domain.Keywords : Image Retrieval, Image Indexing, Image Processing 1. INTRODUCTION Current developments in visual information systems and fast expansion of web sites require flexible, efficient and effectiveimage retrieval schemes. Search for images in databases based on the pattern recognition methods is very difficult. The reason isthe disability of current computer vision schemes in extracting the high level information as human perception can do 3,12 Lowlevel image information such as color, shape and texture can be converted into quantitative factors. These factors can express theimage in different resolutions and are suitable for image retrieval. Several image features can be combined to improve theretrieval effectivity in content based image retrieval schemes (CBIR).Indexing is a suitable method for image retrieval, because it reduces the search space and improves the search time. In imageindexing an N-dimensional vector based on the image features, for each image in the database is generated, and saved with theimage. As this N-dimensional vector represents the image characteristics, it can be used in retrieval process instead of the imageitself. A distance function or a similarity measure is used to find the similarity between two image feature vectors.Two criteria for comparing image retrieval schemes are efficiency and effectivity. These two factors are respectively the speed ofretrieval and the rate of successful search in fmding similar images with the query one. All the image retrieval schemes try toreduce the search space and increase the speed to improve the efficiency and employ rotation, translation and scaling invariantfeatures to improve the effectivity.This paper investigates image retrieval and indexing techniques in pixel domain, which are referred to schemes that the imagefeatures are directly extracted from the image pixels rather than doing any transformation on pixels. There are methods thatextract features from the compressed version of the image, compressed domain indexing