Title of Invention	METHOD FOR PERFORMING CONTENT BASED IMAGE SEARCH
Abstract	The present invention proposes a system and method for content based image search. The present invention describes an algorithm to search an image based on its content wherein CLD descriptors are calculated for each image present in the server database. These image vectors are quantized and are further mapped to different bins depending on the image content. When a query image is entered, a quantized vector is obtained from it which is mapped to one of the bins. Euclidean distance metric is then calculated in order to arrange images within the bin to obtain a mini database which acts as an input for the second level search.

Title of Invention

METHOD FOR PERFORMING CONTENT BASED IMAGE SEARCH

Abstract

The present invention proposes a system and method for content based image search. The present invention describes an algorithm to search an image based on its content wherein CLD descriptors are calculated for each image present in the server database. These image vectors are quantized and are further mapped to different bins depending on the image content. When a query image is entered, a quantized vector is obtained from it which is mapped to one of the bins. Euclidean distance metric is then calculated in order to arrange images within the bin to obtain a mini database which acts as an input for the second level search.

Full Text	FIELD OF INVENTION The present invention in general relates to the field of multimedia applications. More particularly the present invention relates to a system and method for content-based image search. DESCRIPTION OF THE RELATED ART Content based organization of multimedia is one of the emerging applications and of interest to many content service providers. There exist standards such as MPEG7 which try to describe data so that the content in database can be easily archived and retrieved when needed. As the Content is defined by MPEG7, the goal of an efficient and fast content based image search algorithm was identified by three parameters: 1. Accuracy of the result 2. Time of retrieval 3. Scalability A highly efficient algorithm in the existing art gives very good accuracy but the time taken was huge. Thus the need was felt to divide the algorithm in to two stages. The first stage is the elimination stage in which the input is the universal set and the output is highly correlated mini database. The second stage is the selection stage which works on the mini database to give the final set. The present invention involves the first level search only. SUMMARY OF THE INVENTION The main object of the present invention is to develop an Efficient and Fast Content based Image Search Algorithm. The present invention relates to content-based image searching. In particular, the invention describes an algorithm to perform first level search of an image based on its content wherein CLD (Color Layout Descriptor) descriptors are calculated for each image present in the server database. These image vectors are quantized which are further mapped to different bins depending on the image content. Once the query image is entered, a quantized vector is obtained from it which is mapped to one of the bins. Euclidean distance metric is then calculated in order to arrange images within the bin to obtain a mini database which acts as an input for the second level search. Accordingly the invention explains a method for content based image search comprising the steps of: calculating CLD descriptors for each image present in a server database; quantizing and mapping descriptors in different bins depending on the image content; obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search. Routing is done through vector space to reach the bin for each image vector in the database. Query vector is quantized and coefficients of quantized query correspond to particular bin in the space. Calculating Euclidean Distance Metric for Query image vector is done for 9-points space distance. The method further comprising retrieval of similar images from the Query bin and neighboring bins. Accordingly the invention also explains a system for content based image search comprising: means for calculating CLD descriptors for each image present in a server database; means for quantizing and mapping descriptors in different bins depending on the image content; means for obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and means for calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search. These and other objects, features and advantages of the present invention will become more apparent from the ensuing detailed description of the invention taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWING Figure 1 is a flowchart illustrating the concept of the present invention. DETAILED DESCRIPTION OF THE INVENTION The preferred embodiments of the present invention will now be explained with reference to the accompanying drawings. It should be understood however that the disclosed embodiments are merely exemplary of the invention, which may be embodied in various forms. The following description and drawings are not to be construed as limiting the invention and numerous specific details are described to provide a thorough understanding of the present invention/as the basis for the claims and as a basis for teaching one skilled in the art how to make and/or use the invention. However in certain instances, well-known or conventional details are not described in order not to unnecessarily obscure the present invention in detail. The present invention describes an efficient and fast algorithm for content based image search. As the dataset size increases the images list which are mapped to the bin levels also increases. According to an aspect of the present invention, a 9- point CLD descriptor space Euclidean metric concept is introduced for fast image retrieval list from the bin. The Image Descriptors can be visualized as N Dimension Vectors with each vector occupying a specific point in the space. The N dimension vector can be projected into a M dimension Space where M Color Layout Descriptor (CLD) is chosen for distributing the bins. The hashing is done on the coefficients of the descriptor The descriptor has 12 coefficients and the coefficients have different weights attached to it. The M dimensions are determined based on the weight of the coefficient. Each image CLD descriptor coefficient in the database are measured from different 9-point location coordinates by Euclidean distance measure to obtain a 9- Euclidean distance measure and average of all the 9-distance form a eMetric of the image in the respective bin. In query retrieval process, this 9-point space Euclidean Metric concept is applied and matching the list of image eMetric list from corresponding bin list with fine coarse. As shown in figure 1 the Algorithm for first level search can be broken down into the following steps 1. Calculate CLD descriptor for each images in the database (server-based large database). 2. Each image vector in the database is Quantized 3. Routing is done through the vector space to reach the bin for each image vector in the database. 4. Mapping each image vector in the database to the destined bin and dumping in to it. 5. Calculate Euclidean Distance Metric for each image vector in the bin of its distributed coefficient in the space by 9-points space distance 6. Query vector is quantized. Coefficients of Quantized query corresponds to particular bin in the space 7. Calculate Euclidean Distance Metric for Query image vector 9-points space distance. 8. Retrieval of similar images from the Query bin & its neighboring bins. Using 9 Point Euclidian Distance Metric (called eMetric) images in a bin are arranged and the mini database is generated. The time of retrieval is independent upon the size of the universal database and is related to the size of the particular bin. Accordingly the present invention has the following advantages The distribution is unique. The search results in a reduced sample space and the reduction is very large. The combination of the algorithms results in efficient hashing. The accuracy of the algorithm enhances as the size of the database increases and the universal space contains more similar images. Also it is highly scalable, the cost of increasing the size is only increasing the number of bins. The access time is independent of the size of the universal database and is dependent upon the size of a particular bin. It will also be obvious to those skilled in the art that other control methods and apparatuses can be derived from the combinations of the various methods and apparatuses of the present invention as taught by the description and the accompanying drawings and these shall also be considered within the scope of the present invention. Further, description of such combinations and variations is therefore omitted above. It should also be noted that the host for storing the applications include but not limited to a microchip, microprocessor, handheld communication device, computer, rendering device or a multi function device. Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are possible and are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart there from. GLOSSARY OF TERMS AND DEFINITIONS THEREOF Universal Database: The complete database of all the possible images retrieved from the net. Mini Database: The database obtained after applying first level algorithm on the Universal Database. This is the input from second level algorithm. We Claim: 1. A method for content based image search comprising the steps of: calculating CLD descriptors for each image present in a server database; quantizing and mapping descriptors in different bins depending on the image content; obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search. 2. A method as claimed in claim 1 wherein routing is done through vector space to reach the bin for each image vector in the database. 3. A method as claimed in claim 1 wherein query vector is quantized and coefficients of quantized query corresponds to particular bin in the space. 4. A method as claimed in claim 1 wherein calculating Euclidean Distance Metric for Query image vector is done for 9-points space distance. 5. A method as claimed in claim 1 wherein further comprising retrieval of similar images from the Query bin and neighboring bins. 6. A system for content based image search comprising: means for calculating CLD descriptors for each image present in a server database; means for quantizing and mapping descriptors in different bins depending on the image content; means for obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and means for calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search. 7. A method for content based image search substantially described particularly with reference to the accompanying drawings. 8. A system for content based image search substantially described particularly with reference to the accompanying drawings.

Full Text

FIELD OF INVENTION
The present invention in general relates to the field of multimedia applications. More particularly the present invention relates to a system and method for content-based image search.
DESCRIPTION OF THE RELATED ART
Content based organization of multimedia is one of the emerging applications and of interest to many content service providers. There exist standards such as MPEG7 which try to describe data so that the content in database can be easily archived and retrieved when needed. As the Content is defined by MPEG7, the goal of an efficient and fast content based image search algorithm was identified by three parameters:
1. Accuracy of the result
2. Time of retrieval
3. Scalability
A highly efficient algorithm in the existing art gives very good accuracy but the time taken was huge. Thus the need was felt to divide the algorithm in to two stages. The first stage is the elimination stage in which the input is the universal set and the output is highly correlated mini database. The second stage is the selection stage which works on the mini database to give the final set. The present invention involves the first level search only.
SUMMARY OF THE INVENTION
The main object of the present invention is to develop an Efficient and Fast Content based Image Search Algorithm.
The present invention relates to content-based image searching. In particular, the invention describes an algorithm to perform first level search of an image based on its content wherein CLD (Color Layout Descriptor) descriptors are calculated for each image present in the server database. These image vectors are quantized which are further mapped to different bins depending on the image content. Once the query image is entered, a quantized vector is obtained from it which is mapped to one of the bins. Euclidean distance metric is then calculated in order to arrange images within the bin to obtain a mini database which acts as an input for the second level search.
Accordingly the invention explains a method for content based image search comprising the steps of:
calculating CLD descriptors for each image present in a server database;
quantizing and mapping descriptors in different bins depending on the image content;
obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and
calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search.
Routing is done through vector space to reach the bin for each image vector in the database. Query vector is quantized and coefficients of quantized query correspond to particular bin in the space. Calculating Euclidean Distance Metric for Query image vector is done for 9-points space distance. The method further comprising retrieval of similar images from the Query bin and neighboring bins.
Accordingly the invention also explains a system for content based image search comprising:
means for calculating CLD descriptors for each image present in a server database;
means for quantizing and mapping descriptors in different bins depending on the image content;
means for obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and
means for calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search.
These and other objects, features and advantages of the present invention will become more apparent from the ensuing detailed description of the invention taken in conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWING
Figure 1 is a flowchart illustrating the concept of the present invention.
DETAILED DESCRIPTION OF THE INVENTION
The preferred embodiments of the present invention will now be explained with reference to the accompanying drawings. It should be understood however that the disclosed embodiments are merely exemplary of the invention, which may be embodied in various forms. The following description and drawings are not to be construed as limiting the invention and numerous specific details are described to provide a thorough understanding of the present invention/as the basis for the claims and as a basis for teaching one skilled in the art how to make and/or use the invention. However in certain instances, well-known or conventional details are not described in order not to unnecessarily obscure the present invention in detail.
The present invention describes an efficient and fast algorithm for content based image search. As the dataset size increases the images list which are mapped to the bin levels also increases. According to an aspect of the present invention, a 9- point CLD descriptor space Euclidean metric concept is introduced for fast image retrieval list from the bin.
The Image Descriptors can be visualized as N Dimension Vectors with each vector occupying a specific point in the space. The N dimension vector can be projected into a M dimension Space where M Color Layout Descriptor (CLD) is chosen for distributing the bins. The hashing is done on the coefficients of the descriptor The descriptor has 12 coefficients and the coefficients have different weights attached to it. The M dimensions are determined based on the weight of the coefficient.
Each image CLD descriptor coefficient in the database are measured from different 9-point location coordinates by Euclidean distance measure to obtain a 9- Euclidean distance measure and average of all the 9-distance form a eMetric of the image in the respective bin.
In query retrieval process, this 9-point space Euclidean Metric concept is applied and matching the list of image eMetric list from corresponding bin list with fine coarse.
As shown in figure 1 the Algorithm for first level search can be broken down into the following steps
1. Calculate CLD descriptor for each images in the database (server-based large database).
2. Each image vector in the database is Quantized
3. Routing is done through the vector space to reach the bin for each image vector in the database.
4. Mapping each image vector in the database to the destined bin and dumping in to it.
5. Calculate Euclidean Distance Metric for each image vector in the bin of its distributed coefficient in the space by 9-points space distance
6. Query vector is quantized. Coefficients of Quantized query corresponds to particular bin in the space
7. Calculate Euclidean Distance Metric for Query image vector 9-points space distance.
8. Retrieval of similar images from the Query bin & its neighboring bins.
Using 9 Point Euclidian Distance Metric (called eMetric) images in a bin are arranged and the mini database is generated.
The time of retrieval is independent upon the size of the universal database and is related to the size of the particular bin.
Accordingly the present invention has the following advantages
The distribution is unique.
The search results in a reduced sample space and the reduction is very large. The combination of the algorithms results in efficient hashing. The accuracy of the algorithm enhances as the size of the database increases and the universal space contains more similar images.
Also it is highly scalable, the cost of increasing the size is only increasing the number of bins.
The access time is independent of the size of the universal database and is dependent upon the size of a particular bin.
It will also be obvious to those skilled in the art that other control methods and apparatuses can be derived from the combinations of the various methods and apparatuses of the present invention as taught by the description and the accompanying drawings and these shall also be considered within the scope of the present invention. Further, description of such combinations and variations is therefore omitted above. It should also be noted that the host for storing the applications include but not limited to a microchip, microprocessor, handheld communication device, computer, rendering device or a multi function device.
Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are possible and are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart there from.
GLOSSARY OF TERMS AND DEFINITIONS THEREOF
Universal Database: The complete database of all the possible images retrieved from the net.
Mini Database: The database obtained after applying first level algorithm on the Universal Database. This is the input from second level algorithm.

We Claim:
1. A method for content based image search comprising the steps of:
calculating CLD descriptors for each image present in a server database;
quantizing and mapping descriptors in different bins depending on the image content;
obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and
calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search.
2. A method as claimed in claim 1 wherein routing is done through vector space to reach the bin for each image vector in the database.
3. A method as claimed in claim 1 wherein query vector is quantized and coefficients of quantized query corresponds to particular bin in the space.
4. A method as claimed in claim 1 wherein calculating Euclidean Distance Metric for Query image vector is done for 9-points space distance.
5. A method as claimed in claim 1 wherein further comprising retrieval of similar images from the Query bin and neighboring bins.
6. A system for content based image search comprising:
means for calculating CLD descriptors for each image present in a server database;
means for quantizing and mapping descriptors in different bins depending on the image content;
means for obtaining a quantized vector which is mapped to one of the bins once the query image is entered; and
means for calculating Euclidean distance metric to arrange images within the bin to obtain a database which acts as an input for a second level search.
7. A method for content based image search substantially described particularly with reference to the accompanying drawings.
8. A system for content based image search substantially described particularly with reference to the accompanying drawings.

Documents:

2168-CHE-2006 CORRESPONDENCE OTHERS. 04-06-2013.pdf

2168-CHE-2006 EXAMINATION REPORT REPLY RECEIVED 05-03-2013.pdf

2168-CHE-2006 POWER OF ATTORNEY 04-06-2013.pdf

2168-CHE-2006 POWER OF ATTORNEY 05-03-2013.pdf

2168-CHE-2006 AMENDED PAGES OF SPECIFICATION 05-03-2013.pdf

2168-CHE-2006 AMENDED CLAIMS 05-03-2013.pdf

2168-CHE-2006 FORM-1 05-03-2013.pdf

2168-CHE-2006 FORM-13 05-03-2013.pdf

2168-CHE-2006 ABSTRACT.pdf

2168-CHE-2006 CLAIMS.pdf

2168-CHE-2006 CORRESPONDENCE OTHERS.pdf

2168-CHE-2006 DESCRIPTION (COMPLETE).pdf

2168-CHE-2006 DRAWINGS.pdf

2168-CHE-2006 FORM 1.pdf

2168-CHE-2006 FORM 18.pdf

2168-CHE-2006 FORM 5.pdf

2168-CHE-2006 FORM-5 05-03-2013.pdf

2168-che-2006-correspondnece-others.pdf

2168-che-2006-description(provisional).pdf

2168-che-2006-drawings.pdf

2168-che-2006-form 1.pdf

2168-che-2006-form 26.pdf

« Previous Patent

Next Patent »

Patent Number

256378

Indian Patent Application Number

2168/CHE/2006

PG Journal Number

24/2013

Publication Date

14-Jun-2013

Grant Date

07-Jun-2013

Date of Filing

22-Nov-2006

Name of Patentee

SAMSUNG INDIA SOFTWARE OPERATIONS PRIVATE LIMITED

Applicant Address

BAGMANE LAKEVIEW, BLOCK B NO 66/1, BAGMANE TECH PARK, C V RAMAN NAGAR, BYRASANDRA, BANGALORE-560093.

Inventors:

#	Inventor's Name	Inventor's Address
1	BUDDA NAVEEN KUMAR	EMPLOYED AT SAMSUNG INDIA SOFWARE OPERATIONS PVT LTD, HAVING ITS OFFICE AT BAGMANE LAKE VIEW, BLOCK B NO 66/1, BAGMANE TECH PARK, C V RAMAN NAGAR, BYRASANDRA, BANGALORE-560 093.
2	KRITY, KESHAV	EMPLOYED AT SAMSUNG INDIA SOFWARE OPERATIONS PVT LTD, HAVING ITS OFFICE AT BAGMANE LAKE VIEW, BLOCK B NO 66/1, BAGMANE TECH PARK, C V RAMAN NAGAR, BYRASANDRA, BANGALORE-560 093.
3	KALYAN KUMAR KAIPA	EMPLOYED AT SAMSUNG INDIA SOFWARE OPERATIONS PVT LTD, HAVING ITS OFFICE AT BAGMANE LAKE VIEW, BLOCK B NO 66/1, BAGMANE TECH PARK, C V RAMAN NAGAR, BYRASANDRA, BANGALORE-560 093.

PCT International Classification Number

G06F17/30

PCT International Application Number

N/A

PCT International Filing date

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1			NA