HybridDWT-SVD-VQImageCompression forMonochromeImages

Thispaperinvestigates thenew methodof combinedimagecompression fromthewellknown Discrete WaveletTransform (DWT)andSingular Value Decomposition (SVD).The methodmakesuseofthe energycompaction properties ofboththetransforms. On asubblockofanoriginal image2Dwavelet transform is applied. On thetransformed imagethesingular value decomposition iscarried out.Theresultant datacontains maximumenergyoftheimageinaminimumnumberof coefficients. Thesecoefficients alongwiththesingular matrix areVectorQuantized (VQ)and transmitted. Inverse operations arecarried outinordertodecompress thedata. Thusthis paperachieves highcompression byexploiting the energy compaction properties ofthebesttransforms thatare applicable forimagecompression. Simulation results indicate thatthe compression obtainedpossesgood subjective quality ofimage. I.INTRODUCTION Withthedevelopment oftechnology, theproblems ofimage compression haveattracted a lotofattention frommany researchers ofvarious fields like pattern recognition, artificial intelligence, andsignal processing. Imagecompression isa mappingfroma higherdimensional spaceto a lower dimensional space. Allthedigital image compression techniques arebasedontheexploitation ofinformation redundancy that exists inmostdigital images. Theredundancy stemsfromthe statistics oftheimagedatathat isdirectly related totheimage dataprobability distribution andcanbetreated byinformation theory techniques using image entropy concepts. Techniques like Huffman coding, Run-length coding etc. areusedtoremovethe statistical redundancy. Another way ofdescribing image redundancy isby usingPredictive methods. Predictive compression schemes result inlossy compression. Finally, image compression canbeachieved theuseofTransform coding (TC) algorithms. Thepurpose oftransform coding istoconcentrate theimageenergy ina fewtransform coefficients. Transform compression canachieve great data reduction anditiscurrently oneofthebestcoding techniques. Mostpopular transform coding techniques areeither blockbasedorimage- based. Examples ofblock based transform include theKarhunen-Loeve Transform (KLT), Singular value Decomposition (SVD)andthe