Originally posted by nicklevy:
If you know one, post it here! It can be about digital, or standard or both. Here are my two:
1. While the signals from every photocells in a RAW file contain only information about ONE color, and the converter needs information from the surrounding cells to complete the values and make an image, A JPEG file from wich every pixels contain the data from both RGB colors is known to contain much less infromation.
|
JPEG files don't contain every pixel. They contain an approximation of the contents of those pixels. This is why a JPEG file is smaller and also why when you compress them too much, you can see problems.
Originally posted by nicklevy:
2. There are algorithm out there powerfull enough to take ONE BLUE, ONE RED and TWO GREEN information and make four RGB pixels very accurately, but there are none powerfull enough to take 4 complete pixels (with all RGB values) and correctly interpolate the one in between in order to upscale an image without losing quality.
|
Again, not actually true. Look at the edges of tree branches against a sky for example. The purple fringing you see is due to the bayer pattern conversion to RGB pixel values. The purple comes from the increased amount of green photo sites on the sensor, (magenta being the comp colour of green)
The bayer demosaicing has a lot of problems and certainly doesn't give an accurate one to one relationship with an RGB sensor. It just happens to be good enough - though you aren't creating something from nothing.
As for upsampling - you can't create information from nothing - you can just approximate it. That's one of the fundamental laws of physics, not a paradox. cf. with entropy or read anything by Shannon. |
