Hello group,
I am new a DSP novice and am trying to understand the DCT. I have a couple of
questions. My understanding of the DCT is that there are three steps
1) DCT 2) Quantization 3) Compression (zig-zag style like JPEGs).
Does the dct2 function perform all of these functions automatically so or just
the first step?
Some context of what I am trying to do:
Extract features from lip-images for use with visual speech recognition. In the
literature, any lip-images are transformed using the DCT and usually the first
30 or 35 co-efficients taken as input vectors. I am trying to replicate some of
their published experiments and am unclear as to what people mean when they talk
about the "first" n co-efficients. Is it simply the first 30 [i,j] values after
the DCT step or the first 30 co-efficients after the compression step? I
recognise that my query in this regard might be off-topic, but any insights from
DSP experts would be appreciated.
Best regards,
Steve
Does dct2 function perform quantization and compression?
Started by ●June 24, 2009