The coefficients get quantized (rounded to the nearest integer) after dividing by 2^n. That way, each of those numbers occupy n fewer bits on disk. You wanna use a larger n for the high-frequency coefficients since they’re less important. (Except when that assumption breaks, which is fairly often: JPEG compressed screenshots of text is unreadable because text is exclusively composed of sharp high-frequency edges)
Oh and you can figure out what camera/software took a photo by looking at the value of n for each frequency. Everyone does it slightly differently to achieve different compression ratios.
Oh and you can figure out what camera/software took a photo by looking at the value of n for each frequency. Everyone does it slightly differently to achieve different compression ratios.