How audio/video lossy and lossless comprasion works?

[Anonymous543]

hlo there is there anyone here will explain how audio and video compression works?

suppose i have a raw audio file which is recorded at 8 bit per sample/bitdepth and 44,100hz containing only 1 channel
so its bitrate will be around 352.8kbit/s i guess

so if i compress and rencode this file with libopus codec at 128kbit/s
then what data will be removed from raw audio??
bit depth and sample rate will remain same?



for video i dont know the formula of calculating raw video

[jagabo]

Go read some books about information theory, signal processing, and data compression.

Start with:

https://en.wikipedia.org/wiki/Discrete_cosine_transform
https://en.wikipedia.org/wiki/Motion_estimation
https://en.wikipedia.org/wiki/Entropy_encoding

[pandy]

hlo there is there anyone here will explain how audio and video compression works?

Perhaps somebody may be brave enough to start describing coding methods - but this is very wide topic where having strong math background may be necessary to explain priniciples.

then what data will be removed from raw audio??

This is associated with many factors - but mostly lossy audio compression will try to exploit human audio system characteristic - some signals are masked, some signals less audible, some signals not absolutely necessary to be present in reconstructed spectrum.

bit depth and sample rate will remain same?

sample rate yes, bit depth not necessarily

for video i dont know the formula of calculating raw video

bits per pixel (components number * bitdepth) * number pixels * number lines * number frames

[Anonymous543]

hlo there is there anyone here will explain how audio and video compression works?

Perhaps somebody may be brave enough to start describing coding methods - but this is very wide topic where having strong math background may be necessary to explain priniciples.

then what data will be removed from raw audio??

This is associated with many factors - but mostly lossy audio compression will try to exploit human audio system characteristic - some signals are masked, some signals less audible, some signals not absolutely necessary to be present in reconstructed spectrum.

bit depth and sample rate will remain same?

sample rate yes, bit depth not necessarily

for video i dont know the formula of calculating raw video

bits per pixel (components number * bitdepth) * number pixels * number lines * number frames

So video/audio codecs targets that data which humans can't detect easily ,and remove them?

[pandy]

So video/audio codecs targets that data which humans can't detect easily ,and remove them?

This is foundation of the lossy coding - remove information which anyway will be probably not important (normally not perceived) for 99.99% of humans (average listener/watching person).

[Cornucopia]

...Or remove more data, which will be noticeable, but the loss might be considered acceptable for those circumstances.


Scott