« on: July 27, 2014, 05:32:41 PM »
Most of the photo won't have high-resolution components with any real intensity, so it ought to be possible to come up with a hybrid compression scheme that reflects this reality and nets you lots of compression losslessly. I'm just not sure what that compression scheme would look like.
Thats what Sony does with the A7r.
Basically: save the average of a block, and for each pixel in there you store the difference to said avaerage. The difference can use a smaller range of values and thus fewer bits to represent them.
Almost advanced idea: first find out whether or not that will lead to an actual loss of information and flag the block either as standard or compressed.
And important: communicate what happens, otherwise you'll have the cooked raw stigma.