Forum: Pentax DSLR Discussion
07-24-2016, 01:16 PM
|
|
99.99% of world's JPGs are 8-bit DCT encoded pictures and rest are progressive. JPEG has extended version which makes possible to progressively load the image.
The eight bits aren't those familiar eight bits which everyone is so used to. Those are coefficients for cosine function. Now consider the cosine function to output 16 bits which is more accurate representation of the data and allows next processing stage to use those extra bits more efficiently. And every pixel has output touched by many cosine functions. It is the encoder's choice to select quantification coefficients so that image looks good. One trick In JPEG encoders do is they zero the coefficients in high frequencies. Something low as 10 of 64 coefficients makes good enough image.
Any decent image editor should use some internal representation which allows manipulations work with wide enough values. And I believe they all do that. And as stated previously, missing information cannot be restored.
Here are two videos about JPEG which show how the JPEG works. You Tube |
src="https://www.youtube.com/embed/n_uNPbdenRs?controls=1" allowfullscreen> |
You Tube |
src="https://www.youtube.com/embed/Q2aEzeMDHMA?controls=1" allowfullscreen> |
|