Training set in vector quantization -


i learning vector quantization in terms of image compression , reconstruction, wondering why form initial codebook training set? since transmitting index of codeword only, why wouldn't treat training set final codebook. result in more accurate reconstruction, right? difference?

the codebook not image transmitted. codeword reference point row in k-dimensional codebook. note: codebook 2d array row number codeword.

example: 1000 training images of size 512x512 each. if want create codebook sized 256, compress 512x512 image 128x128 vq (vector quantized) index table, our codebook needs 16 dimensions (i.e., each 4x4 block (16 values) in our original image compressed 1 value codeword). whichever clustering algorithm (eg. lbg) use, find closest 16 values best represent 4x4 block in 1000 training images. these 16 values put codebook. after best 256 groups or clusters of 16 values found, codebook ready use.

this codebook needs sent receiver once. 256 row, 16-dimensional structure, compared 1000 512x512 training images. send 128x128 index table receiver can lookup location in codebook , rebuild 512x512 image. not exact, type of compression lossy.


Comments

Popular posts from this blog

Magento/PHP - Get phones on all members in a customer group -

php - .htaccess mod_rewrite for dynamic url which has domain names -

Website Login Issue developed in magento -