I was wondering if anyone at Photokina had the chance to play with the GH2 and got to try out the tele conversion function. It sounds like it's using a full center scan at the maximum telephoto length, which will produce 1080p videos that in theory I guess should look like a 100% crop of a 1920x1080 center of a still image. This is what the Canon T2i and 60D do already with their crop mode, just at VGA resolution.
But what I'm really curious about is how will the camera get from a tele length of 0 to the max of 2.6? Is it just a simple "on" button where it goes straight to the full 2.6 or will it be like most digital zooms and let you zoom in at any length? Or will it have a variety of preset choices, like 1.3, 1.7,2.0,2.6? If it is NOT going straight from no zoom to 2.6 (which is what I'm guessing) then something has to be going on with the camera's binning and scaling routines (unless they're just cropping and scaling the original until you got to the full 2.6, in which case anything in between would look like cr@p). What I suspect is going on is several different sampling routines will be employed along the way as it zooms in more and more, and in between each routine a simple upscaling (or downscaling) process will be used. To illustrate this, I've borrowed how one chip maker (Cypress) performs its subsampling routines and extrapolated it to what might happen in a tele conversion process on one of their chips. If say the original subsampling routine for live view/video mode looks like this:
then as you zoom in to a cropped portion of the sensor you should get subsampling routines that are closer and closer together:
then
then
Finally at some point (where???!) a full sensor scan will be used. If it is before the full 2.6x mode (say it were to kick in at just 2x) then you'd get better quality at 2x since it'd be using a full sensor scan then presumably downsampling to get to 1080. It'd be particularly interesting to know what happens at 1.29x when the cropped portion off the sensor is 3840 pixels wide. I'm guessing there will be certain "sweet spots" along the way where image quality will be optimal for a given routine.
Note the above is NOT how I believe Panasonic implements their video mode - nobody seems to know - but is instead just a simple representation of the general process.
If anyone had a chance to use the feature and inspect the quality of the video as you zoom in I'd love to know their impressions. To me it's the most interesting part of the GH2 and could be more significant than people realize (especially if the full sensor scan mode starts before 2.6x)
But what I'm really curious about is how will the camera get from a tele length of 0 to the max of 2.6? Is it just a simple "on" button where it goes straight to the full 2.6 or will it be like most digital zooms and let you zoom in at any length? Or will it have a variety of preset choices, like 1.3, 1.7,2.0,2.6? If it is NOT going straight from no zoom to 2.6 (which is what I'm guessing) then something has to be going on with the camera's binning and scaling routines (unless they're just cropping and scaling the original until you got to the full 2.6, in which case anything in between would look like cr@p). What I suspect is going on is several different sampling routines will be employed along the way as it zooms in more and more, and in between each routine a simple upscaling (or downscaling) process will be used. To illustrate this, I've borrowed how one chip maker (Cypress) performs its subsampling routines and extrapolated it to what might happen in a tele conversion process on one of their chips. If say the original subsampling routine for live view/video mode looks like this:
then as you zoom in to a cropped portion of the sensor you should get subsampling routines that are closer and closer together:
Finally at some point (where???!) a full sensor scan will be used. If it is before the full 2.6x mode (say it were to kick in at just 2x) then you'd get better quality at 2x since it'd be using a full sensor scan then presumably downsampling to get to 1080. It'd be particularly interesting to know what happens at 1.29x when the cropped portion off the sensor is 3840 pixels wide. I'm guessing there will be certain "sweet spots" along the way where image quality will be optimal for a given routine.
Note the above is NOT how I believe Panasonic implements their video mode - nobody seems to know - but is instead just a simple representation of the general process.
If anyone had a chance to use the feature and inspect the quality of the video as you zoom in I'd love to know their impressions. To me it's the most interesting part of the GH2 and could be more significant than people realize (especially if the full sensor scan mode starts before 2.6x)