[HN Gopher] Nasir Ahmed's digital-compression breakthrough helpe...
       ___________________________________________________________________
        
       Nasir Ahmed's digital-compression breakthrough helped make
       JPEGs/MPEGs possible
        
       Author : Brajeshwar
       Score  : 92 points
       Date   : 2024-08-20 13:52 UTC (9 hours ago)
        
 (HTM) web link (spectrum.ieee.org)
 (TXT) w3m dump (spectrum.ieee.org)
        
       | selimthegrim wrote:
       | His Signals and Systems book is still around and not bad either.
        
       | nayuki wrote:
       | It's about the
       | https://en.wikipedia.org/wiki/Discrete_cosine_transform .
       | 
       | > (subtitle) His digital-compression breakthrough helped make
       | JPEGs and MPEGs possible
       | 
       | Technically, the DCT isn't restricted to only digital
       | compression. The DCT performs a matrix multiplication on a real
       | vector, giving a real vector as output. You can perform a DCT on
       | a finite sequence of analog values if you really wanted to, by
       | performing a specific weighted sum of the values to yield a new
       | sequence of analog values.
        
       | bob1029 wrote:
       | The DCT is really neat, but the actual compression magic comes
       | from a combination of side effects that occur after you apply it:
       | 
       | 1. The DCT (II) packs lower frequency coefficients into the top-
       | left corner of the block.
       | 
       | 2. Quantization helps to zero out many higher frequency
       | coefficients (toward bottom-right corner). This is where your
       | information loss occurs.
       | 
       | 3. Clever zig-zag scanning of the quantized coefficients means
       | that you wind up with long runs of zeroes.
       | 
       | 4. Zig-zag scanned blocks are RLE coded. This is the first form
       | of actual compression.
       | 
       | 5. RLE coded blocks are sent through huffman or arithmetic
       | coding. This is the final form of actual compression (for intra-
       | frame-only/JPEG considerations). Additional compression occurs in
       | MPEG, et. al. with interframe techniques.
        
         | kappi wrote:
         | DCT is now replaced by Hadamard Transform which can be
         | implemented by additions/subtractions and don't have the drift
         | problem of DCT. HT was considered before DCT, but during that
         | time DCT was picked because of better perceptual quality. Later
         | during H.264 standardization, HT replaced DCT and is now used
         | in all video codecs instead of DCT.
        
           | mbtwl wrote:
           | Nope.
           | 
           | X265/HEVC
           | https://en.m.wikipedia.org/wiki/High_Efficiency_Video_Coding
           | 
           | Also not true for X266/VVC.
        
             | aidenn0 wrote:
             | AV1 also uses DCT and DST, but not Hadamard.
        
             | kappi wrote:
             | correct, it is integer DCT. Lot of techniques adopted from
             | the integer transform of H.264. That's what I meant, not
             | the floating point DCT proposed in 70s.
        
           | aidenn0 wrote:
           | Interestingly enough, JPEG XR used a form of the Hadamard
           | Transformation, but JPEG XL (which is newer) uses DCT and
           | Haar transforms.
           | 
           | [edit]
           | 
           | Combined with the information from sibling comments, it seems
           | that the Hadamard transform was something used in standards
           | developed in the '00s but not since.
        
         | pornel wrote:
         | The "actual compression magic" has been used before DCT in
         | other codecs, but applied directly to pixels gave lousy
         | results.
         | 
         | You can also look at 90's software video codecs developed when
         | DCT was still too expensive for video. They had all kinds of
         | approaches to quantization and entropy coding, and they all
         | were a pixelated mess.
         | 
         | DCT is the key ingredient that enabled compression of
         | photographic content.
        
           | HarHarVeryFunny wrote:
           | What's so special about DCT for image compression?
           | 
           | The main idea of lossy image compression is throwing away
           | file detail, which means converting to frequency domain and
           | throwing away high frequency coefficients. Conceptually FFT
           | would work fine for this, so use of DCT instead seems more
           | like an optimization rather than a key component.
        
       | dilippkumar wrote:
       | Hey! Nice to see this here.
       | 
       | My graduate thesis advisor was a coinventor of the DCT [0]. I
       | miss my grad school days - he was a great advisor.
       | 
       | [0]. https://en.wikipedia.org/wiki/K._R._Rao
        
         | bob1029 wrote:
         | I really like the book he co-authored with P. Yip [0]. Grabbed
         | a copy on AbeBooks a few years ago while working on a custom
         | codec. Excellent coverage of the transform from many angles,
         | including reference diagrams of how to implement the various
         | transforms in software/hardware and ~200 pages worth of
         | discussion around applications.
         | 
         | [0]: https://dl.acm.org/doi/10.5555/96810
        
       | trhway wrote:
       | The first layer of the visual cortex (and what the input layers
       | convolutional kernels in visual NN converge to) are those Gabor
       | kernels - cosine multiplied by exponentially decreasing amplitude
       | thus de-facto limiting the spatial attention of the given neuron
       | to a spot.
        
       | max_ wrote:
       | One thing I recommend people to do is study compression
       | algorithms like Jpeg.
       | 
       | I find the relationship between compression algos & cognitive
       | science very interesting.
        
         | drunkspider wrote:
         | What's the relationship between compression algorithms and
         | cognitive science?
        
           | tedd4u wrote:
           | "Lossless" compression is based on information that can be
           | discarded without negative consequences because it cannot be
           | perceived by humans. The data is real and there, you just
           | can't see it or hear it. If you can quantify what information
           | humans can't perceive, you can discard it, leaving less data
           | and possibly more amenable data for a subsequent lossless
           | compression phase. MP3, JPEG, MPEG all benefit from this
           | understanding of the human perceptual system.
        
             | omneity wrote:
             | You're talking about lossy compression. Specifically
             | perceptual lossy compression[0].
             | 
             | Lossless compression is entirely reversible. Nothing is
             | lost and nothing is discarded, perceived or not, like zip.
             | 
             | 0: https://arxiv.org/abs/2106.02782
        
             | hnlmorg wrote:
             | You have it backwards there. You're describing lossy
             | compression.
             | 
             | Lossless is formats like Flac and zip. Lossless compression
             | basically stores the same data in more efficient (from a
             | file size perspective) states rather than discarding stuff
             | that isn't perceived.
             | 
             | The clue is in the name of the term: "lossy" means you lose
             | data. "Lossless" means you don't lose data. So if a zip
             | file was lossy, you'd never be able to decompress it.
             | Whereas you cannot restore data you've lost from an MP3.
        
           | nayuki wrote:
           | Perhaps
           | https://en.wikipedia.org/wiki/Human_visual_system_model ,
           | https://en.wikipedia.org/wiki/Psychoacoustics
        
           | max_ wrote:
           | There are many other resources.
           | 
           | But this is an example https://archive.is/KShWY#9
        
       | laidoffamazon wrote:
       | Extremely impressive, done while doing research at Kansas State
       | University with a PhD from the University of New Mexico. I don't
       | know if any new major advancements have come from people from
       | state schools today.
        
         | mkoubaa wrote:
         | Is this sarcastic?
        
       | duped wrote:
       | This is why it's important to pay attention in linear algebra
       | class as a CS undergrad!
        
       | kleiba wrote:
       | Wikipedia writes: "Ahmed developed a practical DCT algorithm with
       | his PhD students T. Raj Natarajan, Wills Dietrich, and Jeremy
       | Fries, and his friend Dr. K. R. Rao at the University of Texas at
       | Arlington in 1973." [1]
       | 
       | So perhaps it would fair to give due credit to the co-workers as
       | well.
       | 
       | [1] https://en.wikipedia.org/wiki/Discrete_cosine_transform
        
       ___________________________________________________________________
       (page generated 2024-08-20 23:00 UTC)