Does Sony's "clear zoom" work with non-Sony lenses? Or with the LA-EA2 + A-mount?

Started Mar 29, 2014 | Questions thread
captura Forum Pro • Posts: 25,962
Re: Does Sony's "clear zoom" work with non-Sony lenses? Or with the LA-EA2 + A-mount?

Digital Nigel wrote:

GaryW wrote:

Digital Nigel wrote:

....I must admit that, when viewed at equivalent sizes, I can't see any difference between the small ooc JPEG and the much larger CIZ images, but the upsized RAW beats them both, with more detail, fewer artefacts and better colour. There are slightly sharper jaggies visible in diagonal lines in the RAW, but they're only just visible at 100% viewing.

So I wonder what secret sauce Sony uses in its CIZ resampling algorithm that manages to be worse than the pedestrian bicubic?

They did smooth out the jaggies, at least for the obvious lines and curves in the letters, and gave an appearance of more detail in those areas. In the RAW version, these jaggies are still present. In this area, CIZ "won".

It's just that when you looked for actual detail, such as the map shading, it was preserved in the RAW version but somehow smoothed-out in the CIZ version. It seems to me that CIZ should have at least done as well, but it did not. Perhaps this is either an artifact of the JPEG processing or a slight focus difference?

I think the color/exposure problem is unrelated to CIZ, but makes it overall less appealing. In my quick tests, the color was preserved with CIZ (same as JPEG), and using a TC introduced less contrast (along with more detail). So, there are tradeoffs, but I still see CIZ as being a convenience for someone that wants it processed all in the camera. I guess as long as CIZ beats a JPEG crop, it is still useful for JPEG users.

Yes, I think that CIZ should, potentially, be beneficial to JPEG shooters who don't want to do any post-processing. But when I compare the CIZ and the ordinary JPEG at 100%, the CIZ image doesn't look any better -- if anything, it looks worse. You can see it by comparing the first two images in my last post at original size.

The OOC and CIZ JPEGs have identical focus, exposure, colour, etc, as you'd expect. So they look identical at a distance, but the OOC JPEG actually looks better at 100% than the CIZ. That's not really a surprise, as it's purer, without 75% of injected fake pixels muddying up the image.

I also suspect that the same would be true of the RAW, where an image at the original size would looks better than an upscaled one. Again, quadrupling the number of pixels doesn't improve image quality. Of course, there will be differences in colour, exposure, etc between RAW and JPEG, as that's what you try to improve during post-processing. There would be something wrong if the RAW was worse than the JPEG.

I suppose the issue is that there's nearly always a final resampling when a raster image is rendered to a physical device, whether that be a printer or a screen. The result is better if there hasn't been a redundant additional resampling along the way, particularly if that involved heavily adulterating the original 'real' pixels with three times as many fake ones.

So, although CIZ should, in theory, give JPEG shooters something useful, in reality, when you actually examine the results, it's actually slightly worse than just cropping an OOC JPEG. And it certainly gives nothing to RAW shooters, who will get better quality just by cropping as needed during their normal post-processing. And they, too, should avoid the temptation to upscale.

Somebody should perform a similar comparison using a FF camera...the A7 or A7r, using one of the APS-C E-mount lenses...maybe the 24/1.8.

The test would involve comparing use of the 'crop button' on the A7 compared to taking the vignetted image and manually cropping, later. Would the images be of the same quality? (Y/N)

 captura's gear list:captura's gear list
Sony Alpha NEX-7 Canon EOS Rebel SL1 Sony a7 Olympus OM-D E-M10 Canon EF 85mm f/1.8 USM +3 more
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow