One thing to remember is that a RAW file only has 16 (*istD with the 
padded RAW format) or 12 bits per pixel, while the subsequent TIFF will 
have 48 bits per pixel (16 bits per channel). The D's TIFF output is 
only 8 bits per channel (24 bits per pixel). So a 8 bit TIFF should be 
roughly 50$ larger than a D's RAW file and twice as large as the other 
6MP RAW's. The reason that this isn't the case is that the RAW file 
contains more information (it's got all the edge pixels which are 
necessary to produce accurate colour but aren't reflected in the 
resolution of the final image, it also has more EXIF info and an 
embedded JPEG).

A Bayer-pattern sensor interpolates colour data from a group of pixels 
each of which has a colour filter over it (One row of sensor sites will 
be RGRGRGRGRG while the next will be GBGBGBGBGB, yes there are twice as 
many green sites as either other colour, to match the fact that the 
human eye is more green-sensitive than anything else).

-Adam


John Francis wrote:
> On Sat, Nov 18, 2006 at 01:11:05PM -0700, Joseph Tainter wrote:
>> How does one know that a compressed RAW file is actually lossless?
>>
>> Over at dpreview, Steve (apparently the first kid on the block with a 
>> K10D) has posted a DNG file taken at ISO 1600 here:
>>
>> http://www.sendspace.com/file/zsdg52
>>
>> It is a 6 mb file. When you open it in CS2, then resave it as a TIFF, it 
>> becomes a 29 mb file. So clearly the DNG was compressed.
> 
> That doesn't prove it - when I open a (13MB) RAW file from my *ist-D,
> or a 10MB RAW file from a *ist-DS, and save as TIFF, it can take 36MB.
> Even an 8-bit TIFF needs 18MB, which is larger than either RAW file.
> 
> What *does* prove it is compressed, though, is that a 10MP file of
> 12-bit raw values is going to need at least 15MB of uncompressed space.
> If the DNG file only occupies 6MB then it's obviously being compressed.
> 
>> I don't want to use RAW compression unless I cam be completely certain 
>> that it is lossless. How does one know?
> 
> Because the DNG specs say that compressed DNGs use lossless JPEG
> compression, and the JPEG consortium assure us it really is lossless.
> 
> Beyond that, you'd actually need to understand the algorithms used.
> But (unless you're a conspiracy nutcase) you can be certain that a
> compressed DNG isn't throwing away any of your precious data.
> 
> 


-- 
PDML Pentax-Discuss Mail List
[email protected]
http://pdml.net/mailman/listinfo/pdml_pdml.net

Reply via email to