To crop, or not to crop, 1.6x on the 5D!

In your tests the 5D is still a lot sharper. Great job!

Nick_C wrote in post #4402877
Here is a little test I did a while back to see whether it was worth my time carrying a 1.6x body along with my 5D.

My goal whatever it took was to end up with an 8mp image from the 5D with the same fov of the 1.6x cropped body.

As everyone knows you cant just crop a 5D image as the end result is something more closer to 5mp, well I specifically wanted & needed 8mp, no less!

By upsizing using whatever tool I found the best, I managed to upsize the 5D image to approx 22mp, crop to the same fov as the 1.6x image & leaving me with an 8mp image.

Here are my findings, not as shabby as one would think.

#1 Original 5D image:

HTTP response: NOT FOUND | MIME changed to 'image/gif' | Redirected to error image by FLICKR

Now by mounting my 1.6x body on the tripod using the same lens with exactly the same distance from the subject to the front of the lens, I took this next shot, showing the 1.6x advantage of that body.

#2 Original 1.6x Image:

HTTP response: NOT FOUND | MIME changed to 'image/gif' | Redirected to error image by FLICKR

Now by enlarging the original 5D image to approx 22mp, then cropping to as close as I could to the 1.6x image, I ended up with an 8mp image as shown below:

#3 5D Enlarged & Cropped to 1.6x:

HTTP response: NOT FOUND | MIME changed to 'image/gif' | Redirected to error image by FLICKR

Looks very similar when scaled down for the web, but lets have a look again with 100% crops from each image, remember now that one is the native 8mp, the other has been enlarged or blown up if you like, then cropped to give a similar fov, but both images are now 8mp.

#4 100% crop of native 8mp from 1.6x body:

HTTP response: NOT FOUND | MIME changed to 'image/gif' | Redirected to error image by FLICKR

#5 100% crop of enlarged & cropped 5D image, now 8mp the same:

HTTP response: NOT FOUND | MIME changed to 'image/gif' | Redirected to error image by FLICKR

Overall I dont see any major downsides to this technique, certainly no major loss of clarity or jaggies on the images.

So now I just carry my 5D around with me, I dont often need the extra reach of the 1.6x cameras & I certainly dont miss using them at all, but when or if the situation does present itself, I know I can get away with using just the 5D & not carrying another body around with me, dragging myself around is tough enough

Thank you for taking the time to post this, very interesting!

Here's what my eyes see: I see that the 5D 100% crop is sharper, but, the colors of the 100% crop of the 350D appear to be more vibrant. Look at the "4.7" and the hash-like blue lines above the black box that reads "Reference..." The "4.7.." and the blue above the black appears more washed out on the 5D crop. I'm thinking that the differences in the internal image processing is making the difference. Once again, it's a very interesting test, but at the end of the day, I'm thinking that from a distance the crop of the 350D would look better, especially in a print ad.

Thoughts?

Best,

Michael

This was an interesting comparison. I have to also go against the majority and say the 1.6 shot has more detail and better colour, but that would be expected with almost 1/3 more pixels.

However, the 5D came up very well and for occasional cropping as the OP wanted, certainly seems as if it would be more than adequate. After all, we're talking minor differences in 100% crops that could be easily fixed with some post processing, so that prints would probably be indistinguishable for this type of simple contrast picture.

Edit: It might be interesting to see what would happen in photos with more tonal gradations and details, rather than for high contrast simple graphics.

OldA1 wrote in post #4403429
Nice job here - how about just cropping the 5D image without upsizing it first? Considering that the 30D is 8.2mp and a 5D 12.8, 12.8 divided by 1.6 = 8mp wouldn't that be a better way to compare the two? Even though you are losing 0.2mp on the 5D crop, you have maintained the integrity of the image in terms of actual pixels used and eliminate the software modifiying the image? Doesn't the software add interpolated data between pixels when upsizing?

Happy Holidays to all
Peace
Mark

No if you just take the 12.8mp image from the 5D & crop to 1.6x, you end up with barely a 5mp image, not 8mp, you have to start with something much larger in order to crop & come back to 8mp.

RichNY wrote in post #4403454
Nick- You used the same glass on both cameras, right?

Yes, same glass otherwise it would be a completely pointless test.

Dekka wrote in post #4403576
I wonder how this would have gone had you not told which was which and had a poll on which people liked best. Games like that are fun.

Detail looks to be a toss up with this subject, perhaps something with a more intricate design might be more informative.

Nick, is the black in the 5D crop indicative of the black found on the subject? That seems to be the most glaring difference between the two crops. In the 5D, the black looks more composite whereas the black in the native 8MP crop looks more even with less noise. What about the black edging? On the subject, are they crisp lines or does the black bleed ever so much? (I imagine you'd only be able to tell using a magnifying glass, or the like) In the first crop, the black edging isn't too sharp, whereas the upsized 5D appears razor sharp in this regard. Is this a product of the upsizing?

Neat test, regardless.

Thats always hard to tell for sure, but looking at the printing on the tub of discs it looks pretty clear with no bleed.

Yes the extra crispness of the lines are a result of the S-spline method of enlarging, choose bilinear or bicubic & the results are not quite as good.

KrysiaG wrote in post #4403645
Very interesting! Thank you for taking the time to do this and post the pictures.

Your very welcome

M24 wrote in post #4405479
Thank you for taking the time to post this, very interesting!

Here's what my eyes see: I see that the 5D 100% crop is sharper, but, the colors of the 100% crop of the 350D appear to be more vibrant. Look at the "4.7" and the hash-like blue lines above the black box that reads "Reference..." The "4.7.." and the blue above the black appears more washed out on the 5D crop. I'm thinking that the differences in the internal image processing is making the difference. Once again, it's a very interesting test, but at the end of the day, I'm thinking that from a distance the crop of the 350D would look better, especially in a print ad.

Thoughts?

Best,

Michael

Im not sure about that Micheal, I find the sharper the image the better the final print, especially on 30x20" prints, the crop of the 5D enlarged would have the edge over the native 8mp as its more punchy & has better sharpness.

But lets face it, either would be fine, I was just demonstrating that with careful technique you can get away with just a 5D, not a 5D + 30D or whatever, although the cropped bodies do have faster fps so some will no doubt prefer to carry 2 bodies still.

BogongBreeze wrote in post #4405640
This was an interesting comparison. I have to also go against the majority and say the 1.6 shot has more detail and better colour, but that would be expected with almost 1/3 more pixels.

However, the 5D came up very well and for occasional cropping as the OP wanted, certainly seems as if it would be more than adequate. After all, we're talking minor differences in 100% crops that could be easily fixed with some post processing, so that prints would probably be indistinguishable for this type of simple contrast picture.

Edit: It might be interesting to see what would happen in photos with more tonal gradations and details, rather than for high contrast simple graphics.

Remember I made no attempt to match the colours accurately on both images, the main test was to see if there was any "major" downsides to this technique, like jaggies around the edges of the text, blurry images as a result of the enlargement, that kind of thing.

As for another test, I did do exactly the same thing to a real scene (a shot across the street) which consists of trees for fine detail, plenty of things to show up any problems, but the same conclusion was found, the 5D enlarged image was SO close to the native 8mp from the 1.6x body that it was a case of "hmm I cant tell the difference" I got a few people here to have a look & they all picked the 5D enlarged image, saying it had better sharpness!

The main reason I chose a test subject like this was because I have found its easier to see any differences on text rather than say a persons face, also its a controlled enviroment, where as when doing the same test outdoors its hard to get any degree of accuracy in the 2 images, especially at this time of the year with the weather!

what software did you use for this test? CS3?

It's interesting to me that the 350 shows more noise over the 5D with the 100% crops. I see a little more colour with the 350 shot as BongoBreeze points out, but I am putting that down to camera settings.

Overall, glad I have a 5D!

adamlc wrote in post #4405692
what software did you use for this test? CS3?

It's interesting to me that the 350 shows more noise over the 5D with the 100% crops. I see a little more colour with the 350 shot as BongoBreeze points out, but I am putting that down to camera settings.

Overall, glad I have a 5D!

Ive tried everything around & find CS3 to be the worse way of upsizing, photozoom funny enough produced the best results!

It makes sense about the noise as a 350D at ISO100 is still more noisy than a 5D at the same ISO.

Yes as I said to the other guy, I made no attempt to match the colour, could be a result of how the raw's were processed, its mainly to expose any detail loss.

One thing about the colour difference, I dont know if your using an LCD the same as me? but when I see both on screen the top one (350D) has more colour, if I scroll down so its in the same position as where the 5D image was, it looks not so saturated, I put both side by side in photoshop & the differences in the crops in terms of colour are much less, if anything I would say the orange looks deeper on the 5D now..

Bloody LCD's! this one im using is a good one too I like them for their sharpness & colour accuracy, but I sure miss a good old CRT.

I have the Genuine Fractals plugin but never used it yet, I'll check out that Photozoom, thanks.

In this test, it's only the sharpness that matters.... case closed

adamlc wrote in post #4405714
I have the Genuine Fractals plugin but never used it yet, I'll check out that Photozoom, thanks.

In this test, it's only the sharpness that matters.... case closed

Yes I found that the best, has to be the S-Spline method, I never had much success with fractals either, sorry

You can see why I no longer carry a cropped body with me, based on what ive seen I know that IF I want that extra reach I can get away with this technique, its not something I do a lot of as I very rarely miss the 1.6x gain, but its made me realize the differences are very subtle, too subtle to warrant me carting more gear than what is essential to get the job done.

Nick_C wrote in post #4405723
... I never had much success with fractals either, sorry

You git!!

adamlc wrote in post #4405729
You git!!

lol

I was being polite, it completely sucks! still you can always flog it on ebay for a 99p bid im sure