Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Upsample, upscale low-fi audio to improve quality -what tool to use? (Read 16478 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Upsample, upscale low-fi audio to improve quality -what tool to use?

Reply #25 –
Quote from: Arnold B. Krueger on
You are badly mislead if you think that upsampling will provide any kind of practical advantage.

I am aware that e.g. blind spectral band replication must be a very hard problem. Clearly, if we make no assumptions about source and listener, it can be dismissed on information-theoretical grounds (one cannot reliably guess the missing octave of white noise that has been lowpass filtered)

Can you refer me to sources that show subjective listening tests where the perceived advantage was zero?

Thank you.

-k

Upsample, upscale low-fi audio to improve quality -what tool to use?

Reply #26 –
Quote from: mzso on
But upsampling can greatly improve the result, by decreasing artifacts and getting closer to the original signal. Not much fimilar with audio processing, but with images the difference between lanczos and nearest neighbor is drastic.

Audio/video resampling is intimately related to A/D and D/A conversion.

If you had compared an audio D/A converter that was based on closest neighbour to one that was based on state-of-the-art algorithms, you might feel that the difference was significant. Fortunately, audio tends to be better approximations to the sampling theorem.

While image processing and audio processing share many things, there are fundamental differences in how our vision works spatially vs how our hearing works temporally, how cameras works vs how audio ADCs work and how displays work vs how audio DACs work:
-Linear phase is critical for spatially processed images, while it is often of little importance for temporally processed audio
-Our vision seems to use a wide bandwidth spatial transform (if any), leading to behaviour that is most easily understood in the direct spatial domain or using wide-bandwidth frequency transforms (e.g. 8x8 pixel in JPEG). Our hearing seems to use narrow bandwidth temporal transforms, leading to behaviour that is most easily understood in the transformed frequency domain (1000s of samples).
-Image processing scaling is sometimes used to "correct" for flaws in the camera or display, such as lack of sharpness. Audio people tend to prefer controlling their equalizers themselves.

That is not to say that it is impossible to do fantastic things with image scaling, in particular for images that have special, well-defined features. This tends to happen outside of my understanding of regular sampling theory, such as the example below:
http://research.microsoft.com/en-us/um/people/kopf/pixelart/


Note that this image is not a general image. It consists of very sharp (aliased) transitions, very few quantized levels, and the original/hirez reference is assumed to consist of smooth, drawn lines. For this set of conditions, remarkable results are possible.

-k

Upsample, upscale low-fi audio to improve quality -what tool to use?

Reply #27 –
Quote from: knutinh on
Quote from: Arnold B. Krueger on
You are badly mislead if you think that upsampling will provide any kind of practical advantage.

I am aware that e.g. blind spectral band replication must be a very hard problem. Clearly, if we make no assumptions about source and listener, it can be dismissed on information-theoretical grounds (one cannot reliably guess the missing octave of white noise that has been lowpass filtered)

Can you refer me to sources that show subjective listening tests where the perceived advantage was zero?


I've done relevant experiments, but since the results were negative, nothing was published.  I believe my experiments were similar to experiments failing to show the existence of perpetual motion. ;-)

Upsample, upscale low-fi audio to improve quality -what tool to use?

Reply #28 –
Quote from: Arnold B. Krueger on
I've done relevant experiments, but since the results were negative, nothing was published.  I believe my experiments were similar to experiments failing to show the existence of perpetual motion. ;-)

This reminds me of the Systematic review of publication bias in studies on publication bias...