Too many megapixels? Six is enough for an 8x10.

Started Apr 26, 2009 | Discussions thread
VTL Regular Member • Posts: 118
Re: Diminishing returns and the speed / file size issue

Daniel Browning wrote:

The only exception I can think of is cropping in ample light. [...]

This is a good point. For their needs, I don't even think that some "digital zooming" is necessarily a bad thing either, given the relative abundance of resolution (for small prints or on-screen use). Sure, they could crop, but that's not trouble they want to go to. Moreover, there are framing, metering, and focusing advantages.

Instead, since optical zooming has been held up as universally superior, we get super-tele-zooms which are of quite dubious quality at their tele ends. It might well be better in terms of image quality to have slightly more modest zoom ranges combined with digital zooming -- certainly for "average" pictures (not taken at the far tele end) and possibly even when fully zoomed in (possibly with some "digital zooming").

Serious shooters are also allergic to digital zooming, but I wouldn't mind having it in an EVF camera, so long as the RAW file could optionally capture everything (but recording the crop). As I said, there are advantages in framing, metering, and focusing.

Let's address the computer speed problem first, [...]

Of course, computer speed is an issue, but somehow it's not the one which concerns me most. There are certainly software issues (i.e., lack of parallelism), but think the situation has gotten any worse from a relative point of view.

The camera speed and storage problems can be solved with intelligent

Storage issues, along with associated issues such as transmission (across networks, for example) are a bigger issue for me. With the usual RAW formats and high-resolution cameras, hard drives can fill up awfully quickly. Arguably, it hasn't gotten any worse, relatively speaking -- but if one wants to keep one's entire photo library handy (let's say, since we're talking about convenience, on a laptop), it sure has. Without the increased resolution, the growth in storage capacities can keep up with the growth in the photo library of a relatively modest shooter like me. With increased resolution, the situation becomes rather hopeless, at least without better compression.

Unfortunately, the "serious photography" world is wedded to the current, supposedly-lossless RAW formats ("supposedly", since some pre-processing of the supposedly-raw data seems to be becoming the norm). This limits to some large extent how compressible the data is. On the other hand, if one were to accept lossy compression, the situation would become much better storage-wise. I would accept in-camera demosaic-ing and subsequent lossy compression, so long as the format could encode the full dynamic range with sufficient precision. There's currently no good middle ground between JPEG and RAW.

As you said, it's a software issue, in this case also software in the camera (plus any associated hardware acceleration necessary, I suppose). But it's also a problem of mindset -- as long as "serious photographers" insist on fully lossless RAW formats, we'll have to contend with much-larger-than-necessary files.

Camera design should not be limited due to known deficiencies in
software design.

I would argue that software design is now a part of camera design! Unfortunately, the hardware aspect outstrips the software.

A full-frame DSLR with the same pixel density as the 50D is 38.4 MP.
So if 15 MP is enough for the 50D, I would think that 38.4 MP
(15*1.6^2) would be about the same for full frame.

To be a little more conservative and practical, one should point out that one doesn't expect larger-format lenses to provide the same absolute resolution. Admittedly, there are gains in the central part of the image -- which is usually most important anyhow -- beyond what I said. Nonetheless, since people don't typically want to use much larger and much more expensive lenses (cost perhaps being a cubic function of linear size) on full-frame, this somewhat limits the resolution attained on full-frame, especially at the borders.

If one wants to have full chroma
resolution with a Bayer sensor, it's necessary to oversample even

There's no doubt that the actual resolution of images from Bayer sensors is much lower than indicated. I would argue that the lower chroma resolution is no bad thing. Possibly, I could accept a compromise which skews even more heavily in favour of luminance resolution.

What I would really like is for camera and sensor manufacturers to go beyond the basic Bayer sensor. At the present, I would like greater dynamic range far more than greater resolution, for example. Unfortunately, the way the software infrastructure is set up ties us to Bayer sensors. If one were to produce a non-Bayer sensor, one would also have to convince all the software makers to figure out how to demosiac it, under the current RAW paradigm. This is an argument for in-camera demosaic-ing.

I suppose I feel that there's been too much emphasis on resolution. But you're right -- the problem lies in part with the software, and the way the software infrastructure is set up. Probably, part of the reason for the almost purely resolution-based race is that it doesn't require any change in the software infrastructure. Producing a new, higher-resolution sensor is a self-contained advancement (which is made practical by Moore's Law).

Other than workflow issues, I don't object to higher resolutions -- but I'd rather have, in most circumstances, greater dynamic range (for example). I don't think resolution should be the current priority in sensor development. For example, I'd rather Nikon take the D300 and significantly improve its dynamic range while keeping it at 12 MP, rather than moving to 15 MP with no significant improvement in dynamic range. Alas, the latter is much easier for them and much more likely.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow