Content-based image retrieval: anapplication ofMPEG-7 Standard andFuzzyC-Means

Theobjective ofthis study istodevelop and robust tochanges ofthebackground. Texture isaproperty of examine theperformance ofanimage classification system using virtually anysurface, including trees, walls, bricks, hair etc. It FuzzyC-Means (FCM)onalarge setofimages represented by contains important structural information ofsurfaces andwell MPEG-7low-level descriptors. Thisexperimental data setconsistsastheir relationship tothesurrounding environment. offive different categories ofimages. Inaseries ofexperiments we considered 5different categories oftheMPEG-7descriptors related ItISworthstressing thatthequality ofacontent-based tocolors andtextures ofimages. Prior toanyclustering the imageretrleval system strongly depends onthecholce ofthe original spacewasreduced usingthestandard Principal setof low-level visualfeatures. For example,the Component Analysis (PCA). A series ofcarefully organized indoor/outdoor classification (7)canbewellperformed using experiments hasledtoanumber ofinteresting findings astothe global colorhistograms andlocal colordescriptors. On the suitability offuzzy sets intheframework inimage organization and other hand,edgehistograms appear tobeuseful incaseof description, insights into thestructure ofvarious categories and city/landscape classification (8).Forinstance, cityimages their interrelationship. usually contain horizontal andvertical edges. Ingeneral, thedimensionality oftheproblem (viz. the I. INTRODUCTION original feature space) isexcessively high andthis calls forthe Givenanomnipresence ofvisual information, image useofsomereduction techniques. Therearesomeimportant understanding becomes oneofthemostevident challenges. advantages behindspacereduction. At first, clustering Considering theexistence ofvarious archives (say, art algorithms usedinthereduced spaces aremoreeffective In galleries, photograph archives, personalized digital albums,thisstudy weconfine ourselves totheprincipal component etc.) thereisa growingdemandforadvanced query analysis (PCA)(9)regarded asvehicle fordimensionality mechanisms that arecapable ofaddressing perceptual aspectsreduction. ofvisual information. Inorder toaddress theneedsof the Thepaper isarranged inthefollowing manner. Section II users, anumberofcontent-based imageretrieval techniques outlines themethodology usedinthestudy. Section III havebeendeveloped,cf.