GEO SPACIAL MATCHING FOR IMAGE RETRIEVAL.

Geo Spacial Matching for Image Retrieval.

     Abstract

                             Every day the average person with a computer faces a growing flow of multimedia information particularly via the Internet. But this ocean of information would be useless without the ability to manipulate, classify, archive and access them quickly and selectively. While text indexing is ubiquitous, it is often limited, tedious and subjective for describing image content.

                                One of the main problems was the difficulty of locating the desired image in a large and varied collection, while it is perfectly feasible to identify the desired image from a small collection simply by browsing. More effective techniques are needed with collections containing thousands of items.

CONVENTIONAL TECHNIQUES (TEXT ANNOTATION):

                                 To date, image and video storage and retrieval systems have typically relied on human supplied textual annotations to enable indexing and searches. The text-based indexes for large image and video archives are time consuming to create. They necessitate that each image and video scene is analyzed manually by a domain expert so the contents can be described textually. The language-based descriptions, however, can never capture the visual content sufficiently.

                                 For example, a description of the overall semantic content of an image does not include an enumeration of all the objects and their characteristics, which may be of interest later.

                              A content mismatch occurs when the information that the domain expert ascertains from an image differs from the information that the user is interested in. A content mismatch is catastrophic in the sense that little can be done to approximate or recover the omitted annotations. In addition, a language mismatch can occur when the user and the domain expert use different languages or phrases. Because text-based matching provides only hit-or-miss type searching, when the user does not specify the right keywords the desired images are unreachable without examining the entire collection.

CONTENT BASED RETRIEVAL:

                            The problems with text-based access to images have prompted increasing interest in the development of image based solutions. This is more often referred to as Content Based Image Retrieval (CBIR). Content Based Image Retrieval relies on the characterization of primitive features such as colour, shape and texture that can be automatically extracted from the images themselves.

                             Queries to CBIR system are most often expressed as visual exemplars of the type of the image or image attributed being sought. For Example user may submit a sketch, click on the texture pallet, or select a particular shape of interest. This system then identifies those stored images with a high degree of similarity to the requested feature.             

EXISTING SYSTEMS:

                             IBM’s QBIC system is the first commercial CBIR system and probably the best known of all CBIR systems. QBIC supports users to retrieval images by color, shape and texture. QBIC provides several query methods: Simple, Multi-feature and Multi-pass. In the simple method, a query is processed using only one feature. A Multi-feature query involves more than one feature and all features have equal weights during the search. A Multi-pass query uses the output of a previous query as the basis for further refinements. Users can draw and specify color and texture color and texture patterns in desired images. In QBIC, the color similarity is computed by quadratic metric using k-element color histograms and the average colors are used as filters to improve query efficiency. Its shape function retrieves images by shape area, circularity, eccentricity and major axis orientation. Its texture function retrieves images by global coarseness, contrast and directionality features.

The Photo book system (developed at the Massachusetts institute of technology) allows retrieving images by color, shape and texture features. This system provides a set of matching algorithms, including Euclidean, mahalanobis, divergence, vector space angle, histogram, Fourier peak, and wavelet tree distances as distance metrics. In its most recent version, users can define their own matching algorithms.

The I match system allows users to retrieve images by color, texture and shape. I match supports several query similar images: Color similarity, color and shape (Quick), color and Shape (Fuzzy), and color distribution.

PROPOSED SYSTEM:      

                    Currently the most widely used image search engine, the GOOGLE, provides its users with the textual annotation kind of implementation. With lacks of images added to the image database, not many images are annotated with proper description. So many relevant images go unmatched.

                     The most widely accepted content-based image retrieval techniques use the Quadratic Distance and the Integrated Region Matching methods. The Quadratic Distance method, though yields metric distance, is computationally expensive. The conventional Integrated Region Matching is non-metric and hence gives results that are not optimal. Our system uses a modified IRM method which overcomes the disadvantages of both the above mentioned methods. The color feature is extracted using the commonly adopted histogram technique.

                 We also provide an interface where the user can give a query image as an input. The colour feature is automatically extracted from the query image and is compared to the images in the database retrieving the matching images.