sid_19911991
Well-known member
As per his review, R3's image at ISO 12800 looks way better than A1's & R5's image at ISO 6400. A1 image seems to be better than R5 though.
Canon R3 vs Sony A1 chroma noise comparison
- Thread starter sid_19911991
- Start date
If you would like to post, you'll need to register. Note that if you have a BCG store account, you'll need a new, separate account here (we keep the two sites separate for security purposes).
I haven't watched it and don't really care as the R3 isn't a camera I would buy but you have to consider the MP difference of those two cameras.
It's interesting, but at the point he did the comparison at 100% it becomes problematic as an objective comparison. Better would be to view at the same output resolution. Let's see how the r5 downsized to 24 megapixels compares to the r3 at 24 megapixels.
I'll test this once I get the R3. 
Calson
Well-known member
I have photographed wedding receptions outdoors at night with the only light coming from votive candles on the guest tables. ISO 6400 was plenty and most of the time I was shooting at ISO 4000. So I do not test my cameras at settings greater than ISO 6400. I started with shooting chrome were ASA 160 was the best I could get so having digital sensors where even the worst provided ISO 640 as with the D2x was a big improvement.
Shooting as ISO values greater than ISO 400 means effectively loosing 2-bits of color depth so there is a price to be paid in color fidelity. Chroma noise I can fix more readily than problems with the actual colors captured by the sensor and processed into an image file. I first saw this with the Canon 5D camera that provided much lower chroma noise than the Nikon DSLR but where color fidelity and compression resulted in people looking like plastic dolls with the alteration of the tonal range of their skin. This was also true for animals like sea lions where the neck coloration would be distorted.
Shooting as ISO values greater than ISO 400 means effectively loosing 2-bits of color depth so there is a price to be paid in color fidelity. Chroma noise I can fix more readily than problems with the actual colors captured by the sensor and processed into an image file. I first saw this with the Canon 5D camera that provided much lower chroma noise than the Nikon DSLR but where color fidelity and compression resulted in people looking like plastic dolls with the alteration of the tonal range of their skin. This was also true for animals like sea lions where the neck coloration would be distorted.
sid_19911991
Well-known member
So
I personally like the comparison this way. With the R5 & A1, one gets better details but more noise. While the R3 is supposedly way cleaner.
That represents the typical real world use.
I personally like the comparison this way. With the R5 & A1, one gets better details but more noise. While the R3 is supposedly way cleaner.
That represents the typical real world use.
From my view the real world involves some output to a print or to the web rather than the 100% views side by side each one a different size. Either way there is resizing involved to get the same size output. I'd like to view equal size images to compare them. I'll bet the r5 wins then, but we'll see.
sid_19911991
Well-known member
Does it matter if the R5 & A1 win over R3 if they are downsized?
Let's say person X shoots with R3 & 100-500 & person Y with R5 & 100-500 side by side.
Requirements - Subject is a kingfisher in flight. 50 feet away. 30% crop needed.
Assumptions - Semi dull light. So ISO needs to be 4000 for a shutter speed of 1/2500.
In such a scenario R3 easily wins over A1 & R5 don't you think in terms of colour noise performance, which is extremely difficult to correct.
Any one would prefer the R3 right in such a scenario which is quite common, IMO?
Or am I missing something?
Last edited:
sid_19911991
Well-known member
John Navitsky
Well-known member
I would have liked to see the histograms for the respective images. I hypothesize the best way to compare high iso performance is by making the histograms look as similar as possible.
sid_19911991
Well-known member
Yeah, but one would have to assume that the reviewer has exposed them the same way as it is a basic thing to do.
John Navitsky
Well-known member
Yah, especially considering it didn’t seem like he originally intended to do an iso comparison, it seems likely he was trying to match exposure. But it seems like sometimes the idea of using the same iso creeps into noise comparisons.
I guess we'll see how they compare. I think you can't compare or judge until you have two images of similar pixel dimensions side by side. When he looked at both at 100% you could see how much bigger the r5 image was on the screen. When you downsize the 45 mp r5 image to be equal to the 24 mp one it will improve because the downsizing software has more pixels to work with. For every pixel the r3 has on the subject the r5 almost has 2 pixels for the same area. So for example print both at 20x30 inches or make jpegs both 4000x6000 pixels and compare them apples to apples.
For noise comparison, both images should be viewed at the same size. The more you zoom into an image, the more noise becomes evident.
Last edited:
lwan
Well-known member
There is a noise reduction trick that can be pulled with a dual pixels sensor, having two pixels representing one allows to better discriminate image features from noise without altering the ground truth image. Maybe canon oulled that string.
sid_19911991
Well-known member
I am not familiar with this.
Are you saying when you take the pic at 45/50 mp on R5 & A1 & then downsize it, the chroma noise becomes significantly less? Will the ISO performance improve? Dynamic range improve?
sid_19911991
Well-known member
Ok. Are you suggesting that one should always downsize high res/megapixel files to get better noise/iso/dynamic range results?
I believe in general that it is yes to all. Of course downsizing has its own issues, like moire, but in general the interpolation software to reduce the pixel dimensions has more to work with and samples the surrounding pixels so if one is noisy but the neighbors are not it can use that data to result in less noise. Less noise means more dynamuc range. Its a bit counter intuitive, but switch the problem around - what if you had to upsize the 24 mp image to 45 mp? Wouldn' the upsized image have to be worse than the original?
I think a lot in photography has to do with the size of the image, the pixel dimensions, and the viewing distance. To make a fair visual comparison have both be the same size, the same pixel dimensions, and the same viewing distance. By looking at different pixel dijensions at 100% on the screen the one image was bigger and so you could see more of its defects.
arbitrage
Well-known member
You can compare this more easily by using the DPR comparison scene. Either at pixel level or the comparison (downsized) or the print output. And you can download the RAW files if you want to play with different scenarios in your own workflow. DPR has a properly lit scene and a lower light scene.
www.dpreview.com
Studio shot comparison: Digital Photography Review
Expert news, reviews and videos of the latest digital cameras, lenses, accessories, and phones. Get answers to your questions in our photography forums.
www.dpreview.com
sid_19911991
Well-known member
But how does that help as we are purposely comparing 2 different cameras.
At the end of the day I want to know which is nosier or better in the field, a 50 mp R5 or a 24 MP R3.
It doesn't matter what the specifications are.
It helps to compare the same size output from the two different cameras to get a fair comparison. In the real world you want to show someone a photo of a certain size to be viewed at a certain distance. Either here or on Instagram or in a print or a jpeg output of a certain size. Turn both camera images into the same size output and then compare them. The r3 could still be better but you'd have to compare apples to apples which the video didn't.
Diffraction is a good example that illustrates the point. If you look at a diffraction calculator, when viewed at 100% the 45mp camera hits it's diffraction limit after f8 but the 24mp camera doesn't hit it until after f11, so you would conclude the 24mp is better. But when you consider equal 8x10 prints both the 45mp camera and the 24mp camera shows diffraction after f 22. So by comparing only the 100% views side by side you might come to the wrong conclusion that you should buy the 24mp camera because it was better for diffraction.
Same for noise and dynamic range and depth of field and everything else, the best bet is to compare equal size output.
sid_19911991
Well-known member
I get your point. Let me make an analogy to be clear.
Is Tata/Jaguar Range Rover better than say a Suzuki S-cross SUV in terms of performance? It pretty much is except for fuel efficiency.
Now is that a fair comparison? No, as for starters they are vastly different price points etc. But it doesn't matter.
The Tata Range Rover is the better car. Period.
Should one scale down the engine to see which one is faster? If one does that, S-cross might easily win a drag race as it is much lighter.
Same with cameras. Let's say I shoot with an R5/A1 & someone else shoots an identical scene side by side with an R3 & whichever produces the better image in terms of chroma noise & other factors, etc. wins.
Even if you do the comparison test by your method between two different brands, the lens is different.
Example - A 50 1.8 of Canon vs Sony counterpart.
One can always make an excuse the lens of one of the brands was slightly less sharp, thereby not making it an apples to apples comparison of the sensors.
But it doesn't matter right. I am deliberately comparing two different sets of equipment to see which is better.
Doug Herr
Well-known member
It helps to define what makes one better than another, depending on what is being compared. Using the car analogy, is the goal comfort? Safety? Off-road capability? Fuel use? The determination of the 'better' vehicle depends on how it's used and what the objective is.
Same for cameras. Are you comparing chroma noise at a specified image size? At pixel level? The results can and often will depend on what you're interested in. Pixel level noise means little if I'm interested in a 24" x 36" print. If that's the print size I want, I'll compare the two cameras by making a 24" x 36" print from each one.
Though it is an older article, this article does a good job of explaining how downsampling a high megapixel camera works to reduce noise even though when viewed at 100% the lower megapixel camera seems to be less noisy. The last segment is the relevant part for this discussion.
photographylife.com
Benefits of a High Resolution Sensor
What are the benefits of a high resolution sensor? Read this article to understand the true benefit of a high resolution sensor, by Mansurovs
photographylife.com