1. SUMMARY

A system for automated content extraction and database searching, based upon intensity was investigated. The intensity based features were chosen as they are fundamental characteristics of the content of all images, giving this work general application towards database of images from a variety of domain.

REFERENCES

[1] H.Muller, ‘Performance Evaluation in

Content-based Image Retrieval- Overview

and proposals’, University of Geneva.

[2] Manesh Kokare, ‘A survey on current

content based image retrieval methods’,

IETE, 2002.

[3] H.Frigui, ‘Interactive Image Retrieval

using fuzzy sets’, University of Memphis.

[4] Gudivada, V.N. and Raghavan, V.V.

(1995). Content-Based Image Retrieval

Systems, IEEE,

PERFORMANCE EVALUATION

Evaluation of retrieval performance is a crucial problem in content based image retrieval (CBIR). The determination of relevant and non-relevant documents for a given query is one of the most important and time-consuming tasks. Using real users, it takes a long time to judge a large number of documents.

The working definition of relevance: “If you were writing a report on the subject of topic and would use the information contained in the document in the report, then the document is relevant”. Only binary judgments (“relevant” or “non-relevant”) are made and the document is judged relevant if any piece of it is relevant (regardless of how small the piece is in relation to the rest of the document). There is also a need for standardization of evaluation measures, since several measures are slight variations of the same definition. This makes it very hard to compare the performance of systems objectively. To overcome this problem, a set of standard performance measures and a standard image database is needed. After all, the ultimate aim is to measure the usefulness of a system for a user. An overview of existing performance evaluation measures in CBIR is given as follows:

User comparison:

User comparison is an interactive method. The users judge the success of the query directly after the query. It is hard to get a large number of such user comparisons, as they are time-consuming.

Before-after comparison: This is the easiest test method. Users are given two or more different results and are asked to choose the one that is preferred or found to be most relevant to the query. This method needs a base system or, at least, another system for comparison.

Single-valued measures:

Rank of the best match: In this method we measure whether the “most relevant” image is either in the first 50 or in the first 500 images retrieved. 50 represents the number of images returned on the screen and 500 is an estimate of the maximum number of images a user might look at when browsing

Average rank of relevant images:

This method can give a good indication of system performance, although it clearly contains less information than a precision-recall graph. It is vulnerable to outliers, since just one relevant image with a very high rank adversely affects it. A simpler and more robust measure is the rank of the relevant image, which is very useful for CBIR

Precision and Recall:

These are standard measures in the Image Retrieval, which give a good indication of system performance. Either value alone contains insufficient information.

Precision = No. of relevant documents retrieved

Total No. of documents retrieved

Recall = No. of relevant documents retrieved

Total No. of relevant documents in collection

We can always make recall one simply one simply by retrieving all images. Thus precision and recall should either be used together, or the number of images retrieved should be specified. Precision and recall are often averaged, but it is important to know the basis on which this is done

Target Testing:

The target testing approach differs significantly from the other performance measures. Users are given a target image and the number of images that the user needs to examine before finding the target image is recorded. Starting with random images, the user marks images as either relevant or non-relevant.

Error Rate:

It is in fact a single precision value, so it is important to know where the value is measured.

Error rate = No. of non-relevant images retrieved

Total No. of Images retrieved

Retrieval Efficiency:

If the number of images retrieved is lower than or equal to the number of relevant images, this value is precision, otherwise it is the recall of a query. This definition can be misleading since it mixes two standard measures.

Correct and Incorrect Detection:

In this method the number of correct and incorrect classifications is counted. When divided by the number of retrieved images, these measures are equivalent to error rate and precision.

.LIMITATIONS

Of the many modalities of visual features that are possible, color and texture are perhaps the most intuitive for people. However, extraction and representation of color by computer are still very challenging tasks. Images are derived from the projection of the 3-D data onto a 2-D plane. In general computer image analysis provides only limited information about the original 3-D plane. This even complicates the extraction of low-level features. The complete understanding of images by computer is not possible with the current technology. However, in the absence of such understanding automated extraction of properly chosen low-level features greatly enhance operation of image and video storage and retrieval systems

APPLICATIONS

There are many applications where content based image retrieval is important. Some of these are:

1. In Architecture, Real Estate and Interior Design: Allows users for finding similar buildings, decoration of room, which corresponds to more appealing structures from database.

2. In cultural services for exploring museums and art galleries.

3. In education, for example in history, it is useful to have immediate access to images and short video sequences of relevant events and people. In giving a lecture on modern product methods, it is useful to be able to include images of early mass production factories etc.

   
   

4. In film and video archives to find video shots quickly for a particular characteristic such as color, texture, shape or even high level concepts such as particular people, places or objects.

3. In Geographical Information Systems for finding where the local attractions are.

4. In Medicine, the medical literature contains volumes of photographs of normal vs. pathological condition in every part of the body. Diagnosis may require recalling that the current condition resembles a condition from the literature. Since medical treatment is often more defective when given early, it is important to search the literature quickly and accurately.

5. In remote sensing, for example finding which satellite images contain tanks etc.

SYSTEM ARCHITECTURE LAST PART

Image Matching and Multi-dimensional Indexing:

Extracted features of query image are compared with features which are stored in image feature database. To achieve fast

retrieval speed and make the retrieval system scalable to large size image collections an effective multi dimensional indexing is indispensable part of the whole system. The system selects the N images having the greatest overall similarities to the query image.

CONTINUATION OF FEATURE EXTRACTION

Feature extraction plays an important role in content based image retrieval to support for efficient and fast retrieval of similar images from image databases. Significant features must first be extracted from image data. Retrieving images by their content as opposed to external features has become an important operation. A fundamental ingredient for content based image retrieval is the technique used for comparing images. There are two general methods for image comparison: Intensity based (color and texture) and Geometry based (shape).

SYSTEM ARCHITECTURE

The basic idea behind content based image retrieval is that, when building an image data base, or retrieving an image from the database, we first extract feature vectors from images (the features can be color, shape, texture, region or spatial features, features in some compressed domain etc.), then store vectors in another database for future use. When given a query image, we similarly extract its feature vectors and match these vectors with those already in the database. If the distance between two image feature vectors is small enough, we consider the corresponding image in the data base match the query. The search is usually based on similarity rather that on exact match and the retrieval results are then ranked according to a similarity index. Usually, a group of similar target images are presented to users. In general, content based image and video

database systems require the following components:

1. Identification and utilization of intuitive visual components.

2. Effective feature representation and discrimination.

3. Automatic extraction of spatially localized features.

4. Techniques for efficient indexing.

The block diagram consists of following main blocks ­–

Digitizer:

To add new images in image database or query images which are acquired from CCD camera, X-ray imaging system, Micro densitometers, Image Dissectors, Video cameras are needed to be digitized, so that the computer can process those images.

Image Database:

The comparison between query image and images from image database can be done

directly pixel by pixel which give precise match but on the other hand, recognizing objects entirely at query time will limit the retrieval speed of the system, due to the high expense of such computing. This crude method of comparison is not used, but image database, which contains raw image, is required for visual display purpose.

Feature Extraction:

To avoid above problem of pixel by pixel comparison, next abstraction level for representing images is the feature level. Every image is characterized by a set of features such as texture, color, shape and others. Extract these features at the time of injecting a new image in image database. Then summarize these features in a reduced set of K indexes and store it in Image Feature Database. The query image is processed in the same way in the database. Matching is carried out on feature database.