Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!

SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

ep1:is lower resolution better--
View previous topic :: View next topic  


PostPosted: Mon Feb 22, 2010 5:25 pm    Post subject: ep1:is lower resolution better-- Reply with quote

or am i crazy (or both) Laughing

i did a little informal test today because i have been upset with the amount of noise i have seen in my ep1 photos at 100% crop. before today i had taken raw + jpeg at highest resolution (12mp). today i shot some raw + medium resolution (8mp), still lowest compression setting. again, this is informal, but to my eye the noise was less than i had been getting and pictures had more clarity at the 8mp setting! am i looney or is it possible that the sensor is too small to accomodate 12mp?

any comments appreciated! (as long as theyre not too nasty... Very Happy


PostPosted: Mon Feb 22, 2010 5:51 pm    Post subject: Reply with quote

The native res of th sensor is 12mp. I suspect that the 8mp image would be similar to the result of downsizing the 12mp image to 8mp using PC/MAC software.

I do agree that 12mp is a bit pushed. I'd rather have 8mp or even a 6mp u4/3 sensor.


PostPosted: Mon Feb 22, 2010 6:07 pm    Post subject: Reply with quote

B: what does that really mean, 'the native resolution of the sensor is 12mp'? does it mean the ep1 sensor is made to accomodate 12mp despite it's size, or is the sensor the same size as it would be/is on a larger camera, or that olympus 'says' it's made for 12mp, but really they just stuffed a whole bunch of pixels into a sensor too small to optimize them?


PostPosted: Mon Feb 22, 2010 8:37 pm    Post subject: Reply with quote

rbelyell wrote:
B: what does that really mean, 'the native resolution of the sensor is 12mp'? does it mean the ep1 sensor is made to accomodate 12mp despite it's size, or is the sensor the same size as it would be/is on a larger camera, or that olympus 'says' it's made for 12mp, but really they just stuffed a whole bunch of pixels into a sensor too small to optimize them?


The sensor is an array of photo pixel sensors. Each pixel sensor consist of a photo-detector and an active amplifier. There're a finite and exact number of pixel sensors on the e-p1's MOS sensor. It's a little bit more than 12000. The number of pixels is built into the cam and not some thing one can change with the menu. I believe when you choose to shoot 8mp, the cam just down-scales the jpeg. I'm pretty sure that the raw file would still be 12mp.


PostPosted: Mon Feb 22, 2010 9:08 pm    Post subject: Reply with quote

yes, that is true; the raw file i believe is still 12mp. my curiosity stems from whether this size sensor--whatever it is designed to fit--can optimize 12mp the way a larger sensor can. for example, my 5d i believe has a physically larger sensor so the pixels are not 'packed so tight', if that is a correct term. i have read before that if you pack too many pixels into a physically small space you get noise--i thought this was one of the problems of the 'pixel wars'...


PostPosted: Mon Feb 22, 2010 9:36 pm    Post subject: Reply with quote

this is the sensor size of your cams
s90-ep1-5D
the ep1 sensor is little bigger than the s90
trying to match the ep1 and 5D is not a fair game


PostPosted: Mon Feb 22, 2010 9:39 pm    Post subject: Reply with quote

rbelyell wrote:
i have read before that if you pack too many pixels into a physically small space you get noise--i thought this was one of the problems of the 'pixel wars'...


When compared at 100%, this is the case, but as you note in your original post, downsampling the image reduces the appearance of the noise as well; this is why the 8 MP image looks better. Also one reason why noise comparisons should be made at a common size (e.g. print or resized on screen) instead of 100%.

In practice you will have more versatility keeping the image at 12 MP instead of downsampling immediately. You can later downsample directly to the final display size (or print the image, which should achieve the same).


To combat the noise itself, you could try overexposing your images somewhat and bringing them back down in raw processing; this should reduce noise (which is more prominent in the shadows), but of course there's a slight risk of clipping highlights in the raw.


PostPosted: Mon Feb 22, 2010 10:51 pm    Post subject: Reply with quote

poilu wrote:
this is the sensor size of your cams
s90-ep1-5D
the ep1 sensor is little bigger than the s90
trying to match the ep1 and 5D is not a fair game


yes, that was my point exactly! the 5d sensor is so much bigger that i cannot understand how the ep1 sensor can pack MORE pixels in without degrading the quality of the image--which is what i was seeing with ep1 raw and 12mp jpg images. is it not true that 'overpacking' (i am sure i am using incorrect terminology) of pixels onto a small sensor can cause increase in noise in photos?


PostPosted: Mon Feb 22, 2010 10:56 pm    Post subject: Reply with quote

Arkku wrote:
To combat the noise itself, you could try overexposing your images somewhat and bringing them back down in raw processing; this should reduce noise (which is more prominent in the shadows), but of course there's a slight risk of clipping highlights in the raw.


thank you arkku; i actually tried that, but i find the ep1 easily blows out highlights that i routinely recapture on my 5d. optimally, i find i need to underexpose a scene, then push in PP, in order to maintain highlights (and that's shooting raw)

i ask you as well, as someone who knows much more about this than i, can a sensor be too small to optimally support a certain number of pixels. in other words, does the physical size of the sensor itself limit the number of pixels it can optimize before inordinate noise sets in as a result of pixel blurring?


PostPosted: Tue Feb 23, 2010 7:53 am    Post subject: Reply with quote

rbelyell wrote:
can a sensor be too small to optimally support a certain number of pixels. in other words, does the physical size of the sensor itself limit the number of pixels it can optimize before inordinate noise sets in as a result of pixel blurring?


With current sensor technology, as far as I know, the smaller the sensor, the more the noise. Hence point and shoots generally have terrible high ISO and your 8MP camera phone still produces crappy results at 100% crop Smile


PostPosted: Tue Feb 23, 2010 9:14 am    Post subject: Reply with quote

the noise is not a problem
you can apply filter and remove it
the problem is that you remove also most details
here is a sample of the ep-1 from dpreview at 200 iso
you don't see noise, but you don't see any details
it is because even at 200 iso, the noise reduction is already extreme
compare this with a 100% crop of Orio carnavale series

http://www.dpreview.com/galleries/reviewsamples/albums/olympus-e-p1-review-samples/slideshow


PostPosted: Tue Feb 23, 2010 12:19 pm    Post subject: Reply with quote

poilu wrote:
the noise is not a problem
you can apply filter and remove it
the problem is that you remove also most details
here is a sample of the ep-1 from dpreview at 200 iso
you don't see noise, but you don't see any details
it is because even at 200 iso, the noise reduction is already extreme
compare this with a 100% crop of Orio carnavale series


Was that really the results of noise-reduction? It might be diffraction in play, IMVHO (aperture = f13). I remember reading in most reviews that the kit lens got drastically soft at f11 likely due to the high pixel density of the m4/3 sensor.

I do agree that for picture quality (noise in this discussion), no one with sane mind would choose m4/3 over a 5d.


PostPosted: Tue Feb 23, 2010 1:02 pm    Post subject: Reply with quote

bogolisk wrote:
poilu wrote:
the noise is not a problem
you can apply filter and remove it
the problem is that you remove also most details
here is a sample of the ep-1 from dpreview at 200 iso
you don't see noise, but you don't see any details
it is because even at 200 iso, the noise reduction is already extreme
compare this with a 100% crop of Orio carnavale series


Was that really the results of noise-reduction? It might be diffraction in play, IMVHO (aperture = f13). I remember reading in most reviews that the kit lens got drastically soft at f11 likely due to the high pixel density of the m4/3 sensor.

I do agree that for picture quality (noise in this discussion), no one with sane mind would choose m4/3 over a 5d.


right, F13 is probably to much for this sensor
I downloaded this sample as I wanted to see people face quality
with those nice models, I thought it was a professional shot; I was wrong


PostPosted: Tue Feb 23, 2010 1:20 pm    Post subject: Reply with quote

so is the conclusion that the ep1 sensor is too small to optimize the number of pixels they chose to stack it with? i totally understand the ep1 vs 5d picture quality is not comparable; but at $6-700 (and $1000usd for ep2 w same sensor) i was not expecting such obvious noise or the extreme detail/noise tradeoff in fairly good shooting conditions. i was only using the 5d as an example of a much larger sensor encompassing less mp's than the ep1, and thus that situation leading inevitably to more obvious noise...

would the IQ have been better had olly chose to optimize this size sensor for 8mp vs 12mp?

btw-wouldnt the only reason to shoot models at f13 be to acheive 'soft' focus?


PostPosted: Tue Feb 23, 2010 3:02 pm    Post subject: Reply with quote

rbelyell wrote:

i ask you as well, as someone who knows much more about this than i, can a sensor be too small to optimally support a certain number of pixels. in other words, does the physical size of the sensor itself limit the number of pixels it can optimize before inordinate noise sets in as a result of pixel blurring?


Well, the noise itself is not really such an issue at the pixel level; as I said above, it should not be assessed at 100% but rather at the desired output size (independent of the pixel resolution).

However, diffraction imposes a physical limit on the maximum resolution of any optical system, and it depends on the f-stop (i.e. physical size of the aperture relative to the focal length). The more you stop down, the lower the maximum resolution you can get—once you stop down beyond the sensor's limit, the extra pixels won't help you get any more resolution and the image will look soft at 100%.

But note that this exact same limit also affects film and indeed any optical system—it has nothing to do with pixels, the pixels just impose another limit. We can therefore question the sense in cramming so many pixels to a small sensor that diffraction prevents the limit of the pixel resolution from being achieved in many (most? all?) situations—this is probably the case with some cameraphones and such already.

One should remember, though, that pixels themselves aren't “evil”, and indeed there are some reasons why it may even be good to put in (slightly) more pixels than would seem sensible in terms of diffraction. For example, most cameras use colour filter array sensor where each “pixel” only “sees” one colour; having more than the “sensible” number of pixels in a given area would therefore give better colour information even though the image itself would still have a soft appearance at 100%—but again, 100% is not in itself a sensible way to view or assess digital images. Actually pixel peeping reviewers (e.g. dpreview) are, in my opinion, very harmful for the progress of camera technology because they entice camera manufacturers to optimise their camera design for 100% viewing rather than realistic images.


PostPosted: Tue Feb 23, 2010 3:07 pm    Post subject: Reply with quote

rbelyell wrote:
so is the conclusion that the ep1 sensor is too small to optimize the number of pixels they chose to stack it with?


IMO yes. btw, it's from Panasonic and it's the same sensor as in the G1 or the GF1. 8MP would have been better ...for me that is.

rbelyell wrote:

i totally understand the ep1 vs 5d picture quality is not comparable; but at $6-700 (and $1000usd for ep2 w same sensor) i was not expecting such obvious noise or the extreme detail/noise tradeoff in fairly good shooting conditions. i was only using the 5d as an example of a much larger sensor encompassing less mp's than the ep1, and thus that situation leading inevitably to more obvious noise...


The e-p1 (along with the GF1, E-P2) are the first EVIL non-dslr-bump cameras and we (you and I) pay a premium for being early adopters. In my case, I can bring it in events where the dummy security guard would confiscate all camera with a dslr-bump (fake or not.) I even told one guy my e-p1 (with the pen-f 38mm) is an old film camera and he believed it.

Now back to the IQ topic, dpreview's review show that the e-p1's output, in not too dark settings, is comparable to the Nikon D5000 and the Canon 500D.
from dpreview's review:
Quote:

...and at a pixel level the E-P1 produces, sharper, crisper output than the Nikon...there's no denying that the little Olympus is giving the D5000 a good run for its money... the little Olympus has the edge when it comes to pixel-level sharpness - despite the Canon's higher nominal resolution...That all said, credit where it's due; the E-P1 is almost matching the 500D's resolution despite the megapixel gap, and the EP-1's pixel level sharpness is noticeably better


So you should compare the e-p1 to entry-level Canikon DSLRs and it would pretty comparable. Against a FF, not even remotely close.


Last edited by bogolisk on Tue Feb 23, 2010 3:13 pm; edited 2 times in total


PostPosted: Tue Feb 23, 2010 3:10 pm    Post subject: Reply with quote

rbelyell wrote:

would the IQ have been better had olly chose to optimize this size sensor for 8mp vs 12mp?


With an 8 MPix sensor, images from an otherwise identical camera would look better at 100% but possibly (though not necessarily) worse at a given realistic output size. However, if Olympus had optimised their 12 MPix sensor design (anti-aliasing filter, A/D conversion, etc) with the idea that all of the images would be downsampled to 8 MPix, it would probably be the best in most cases… (Now you have more versatility, though, since you can assess the quality on an image-by-image basis.)


By the way, if you are a Mac user, try Raw Photo Processor's raw conversion in “half” mode. This will give you only half of the pixel dimensions (i.e. a 4 MPix image from a 12 MPix sensor), but you will effectively get “full colour” pixels instead of the usual interpolation.


PostPosted: Tue Feb 23, 2010 3:16 pm    Post subject: Reply with quote

Arkku wrote:
Actually pixel peeping reviewers (e.g. dpreview) are, in my opinion, very harmful for the progress of camera technology because they entice camera manufacturers to optimise their camera design for 100% viewing rather than realistic images.


OTOH, if reviewers are all like Luis, who keeps cranking out masterpieces using crappiest cameras/lenses, then manufacturer would just produce crappy cam and lens Twisted Evil


PostPosted: Tue Feb 23, 2010 4:07 pm    Post subject: Reply with quote

@bogo: yeah, i agree also. do you have also the problem of easily clipped highlights? i find i have to shoot maybe up to a stop down to make sure i capture sky highlights...easier to recapture shadow detail as recapturing highlight detail is almost impossible...

@arkku: well for now, im trying to get best of both worlds by shooting raw + 8mp jpg and analyse on picture by picture basis. so far, w small sample, i havent used one raw yet. also, thank you so much for the RPP site-if i can download sample copy i will definitely give it a try. 2 questions if you wouldnt mind the help: how is it that 4mp is half size of 12mp? and if RPP half will develop raw at 4mp, i assume the resolution/sharpness will be greatly effected?
thanks again!


PostPosted: Tue Feb 23, 2010 4:20 pm    Post subject: Reply with quote

rbelyell wrote:
@bogo: yeah, i agree also. do you have also the problem of easily clipped highlights? i find i have to shoot maybe up to a stop down to make sure i capture sky highlights...easier to recapture shadow detail as recapturing highlight detail is almost impossible...


Haven't notice that. It's been gray and snowing around here since I got my e-p1 (Nov 09). I've been shooting mostly indoor portraits. Maybe I'll see it when spring come and I start shooting outdoor.

One of the issue is the live histogram reflect the JPEG output which depends on the selected WB and picture mode. I heard that if you set the picture mode to muted or portrait, contrast to -2, then the live histogram would be closer to the raw values. Of course, you'd have to throw the jpeg away and only use the RAW file.


PostPosted: Tue Feb 23, 2010 8:14 pm    Post subject: Reply with quote

rbelyell wrote:

@arkku: well for now, im trying to get best of both worlds by shooting raw + 8mp jpg and analyse on picture by picture basis. so far, w small sample, i havent used one raw yet.


How about if you take the raw and resample <em>that</em> to 8 MP. You could apply slightly stronger noise reduction on it while at 12 MP (since you can afford to lose 4 MP worth of detail, such that it was in the first place) and then downsample + sharpen.

rbelyell wrote:

also, thank you so much for the RPP site-if i can download sample copy i will definitely give it a try. 2 questions if you wouldnt mind the help: how is it that 4mp is half size of 12mp? and if RPP half will develop raw at 4mp, i assume the resolution/sharpness will be greatly effected?!


You're right, the 4 MP was a math error—computer scientists should not attempt this kind of advanced arithmetic without a calculator. =)

The correct figure is of course 3 MP; halve each dimension and you get one fourth of the pixels (4000 × 3000 = 12 M, 2000 × 1500 = 3 M). This corresponds to the “50%” zoom of the original image.

3 MP may be too small to be of much practical interest for anything but web use, but I still recommend trying it, as it shows how much better the “100%” view looks after we take traditional demosaicking out of the picture and simulate actual RGB (or GRGB) pixels. Might be that in the future we'll have cameras intended for this kind of processing…