> '& \pJava Excel API v2.6 Ba==h\:#8X@"1Arial1Arial1Arial1Arial + ) , * `'DC, 4#titletitle[alternative]title[translated]author[google(nobreakline)] contributor[author(nobreakline)]"contributor[scopusid(nobreakline)]date[issued]relation[journaltitle]identifier[issn] citationsidentifier[major]subject(nobreakline) publisheridentifier[thesisdegree]contributor[advisor]relation[ispartofseries]relation[index(nobreakline)]typeabstractidentifier[govdoc]identifier[uri]identifier[doi(doi)]identifier[isbn]identifier[ismn]
identifierפ D tǩ\ ) \ l3content based image search based on color histogram1997YY YtTŐYP YYMasterMaster's ThesisMThe rapidly changing telecommunication technique and wide spread use of internet have enabled us to easily access a lot of information.
And the development of digital library has made it possible to search data in real time even though a database is located in a remoted place.
However, an effective search method for multimedia database with various data has not been developed yet, and the current search method of multimedia database and digital library has limited capabilities. We have studied search methods for the multimedia database.
This paper presents an improved algorithm for a content based search, which is one of the image search algorithms. The majority of existing algorithms for image search are based on color distribution of images. They calculate a color distribution of query images in a histogram, compare it with a color histogram of images in a database and seek highest match value.
However, color histogram used in the algorithm shows color distributions only and does not have any consideration of distance between colors distributed at all. Therefore, it can miss the true image due to minor color changes.
We can see many works on effective image searches considering the distance factor. The proposed algorithm is based on Histogram Intersection of Swain. It tries to enhance search effectiveness by using not only color histogram showing color distribution of images but also inverse covariance matrix, a new weighting factor of the color distance.
The existing algorithm always provides the same number of bins for each color component for calculating color histograms. However we have done computer simulations on the relationship between the number of histogram bins and search effectiveness by testing cases where the number of histogram bins are differently divided based on color distributions.
The image search using Inverse Covariance Matrix as suggested in this paper showed enhanced efficiency in image search of Swain or other algorithms considering the distance factor. Moreover, appropriate quantization of histogram bins has enable us to get more accurate and faster image search results. We accomplished efficient search in terms of time and accuracy by increasing the number of bins for color components by reducing numbers for color components with small variations.; ĳ\ XՔ 0 x07X T <\ ι@ | } ` ǌ . 8 |t췬 t <\ @ 8 Ǵĳ ȹX XX pt0ѠtǤ| XՔ ٳ t t L.
췘 DLɔ t 䲑\ pt0| ǔ @ pt0ѠtǤ \ (x )t , ֬ @ pt0ѠtǤ@ 8 |t췬X @ 买 \x t . t @ pt0ѠtǤX ) \ ι@ l ɉ <ǘ Ɛ ̹q` ̹\ | X .
|8 )X Xx 0D T(content based search) ) t \ LବD Ht .
0t ǔ X Lବ@ X 0D T Lବ<\, x(query image)X | פ <\ и pt0ѠtǤ´ ǔ X פ DPX @ i (match value)D ɔ D >̔ )t.
췘 t Lବ XՔ פ Ŕ X ̹D д , tX p \ $ <\ \ x ||ĳ Ǭ\ | t ٳ|\ פ D \ ٳ|\ <\ xXՔ \Ĭ| .
t pƌ| $\ ( \ ι@ l ɉ . |8X Lବ@ SwainX Histogram Intersection D T <\ X | д פ п DȲ| pƌ\ \ ƌx ,D X (D tଐ X.
, 0t Lବ D פ <\ м L @ 4XՌ פ lD m |XՌ . 췘 פ X lD 0| t D X LѤ¸| X פ X l @ (ǬtX Ĭ tĳ lt X.
|8 HXՔ ,D X D \ 0tX SwainX Lବtǘ x p ƌ| tǩ\ Lବĳ Lt ĳ @ <b (D .
l | XՌ ŐTXՔ @ `t UXՌ D Xĳ] \䲔 D L . T p 1@ l | T ιt 8 X, T @ 1@ lЌ lD U1X !t (x t t踴ĳ] X.~http://dspace.ewha.ac.kr/handle/2015.oak/199232;
http://dcollection.ewha.ac.kr/jsp/common/DcLoOrgPer.jsp?sItemId=000000026122;@Ce|(!CPrs BKm~ AJlmLnWyvi%g
F(^+
dMbP?_*+%" ,,??U
!
"
l>@
!"#$%Root EntryWorkbook,SummaryInformation(DocumentSummaryInformation8