Sunday 6 January 2013

RAW conversion in DPP vs Aperture 3

I'm trying to decide if I should switch from using DPP to Aperture 3. Aperture would almost certainly be easier to use so the only reason not to switch would be if DPP is better at processing images. I've assumed that it would be (the "only Canon knows how to read Canon RAW files" argument) but I figured it was time to check. 

Raw Conversion

The first thing I looked at was how DPP and Aperture interpret the same RAW file. 

RAW file as seen in DPP:



RAW file as seen in Aperture:

These things are always subjective but, to my eye, the rendition in DPP is better. Colours and skin tone are more natural. 

Of course, the whole point of imaging software is to manipulate the image so maybe it is unfair to just look at the RAW file. In DPP, I apply Canon Portrait style and I get:
 In Aperture, I use the Skin Tone eye dropper on her forehead to adjust the colouring.
The colours in the clothes have more pop in the DPP version, similar to what we see in Aperture, but I still find the DPP rendition more natural. 

Interestingly, in Aperture the red channel goes off the scale. Not sure if this is the same as blowing the highlights or if it is just a difference in scale:

The other major difference between DPP and Aperture is the brightness. The DPP rendition is slightly darker, which makes for a more muddy image.



Adjusting Brightness in DPP by 0.33 lights up the face more while keeping the colours from blowing out. 



That gives me an image that, I think, preserves the nice skin tone while matching the brightness of Aperture:

DPP RAW file, Portrait style, Brightness +0.33

Aperture RAW file, Skin Tone White Balance

I am sure that, given the powerful features available in Aperture, it would be possible to adjust the above image to get what I am looking for. But I don't really want to have to go through a major operation for every photo. DPP gives me a pleasing image to start with and then, if I want, I can apply more complicated adjustments. 

Printing

I print my photos so it's not enough that they look good on screen, I also need them to look good on paper. With that in mind, I printed both these images and scanned them (on the same Canon PIXMA MP610) so I can compare:


There lots going on here. First of all, the printed photos look different to the on-screen image. That's to be expected, pixels of light and blobs of ink work very differently and it's not easy to match one to the other, especially on a cheap consumer printer. That said, I don't understand why the difference is so much greater in the case of Aperture. 

Screenshot DPP RAW file, Portrait style, Brightness +0.33
Scan of DPP RAW file print

Making allowances for the fact that printing and rescanning introduces several new variables that need to be controlled, the above images are not a bad match. 

Screenshot of Aperture RAW file, Skin Tone White Balance




Scan of Aperture RAW file print
These images look very different. And it's not the scanner - the printed photograph does look more yellow. 

I didn't know what was going on here. I figured there must be something wrong with the printer settings in Aperture but I cannot see what.

Printer Settings

Canon's ICC Profiles Guide to the rescue. Canon recommends switching off Black Point Compensation when using their custom profiles (which I am doing: GL2 for Photo Paper Pro Plus Glossy II) and, reading between the lines, Perceptual might be better suited to Portraits than Relative Colormetric when setting the Rendering Intent. For an explanation of the difference between the two, see here. Alternatively, I like the analogy from Michael at Luminous Landscape.

Printing from DPP, I use Easy Print, which doesn't give the option to turn off Black Point Compensation. Presumably, Canon takes care of that. I do have the option to switch from Relative Colormetric to Perceptual. Canon recommends Relative Colormetric when working with Easy Print but I cannot really see that it makes a huge difference except possible that Relative Colormetric shows more detail (more freckles.)




Aperture is more interesting. Printing from there, I get both options and there is an effect. 

To start with, Perceptual looks brighter than Relative Colormetric:



Black Point Compensation is more subtle. I'm guessing it is important, since Canon specifically recommends you turn it off, but it's hard to see the impact. With Perceptual, I cannot see any real difference. With Relative Colorimetric, there may be slightly more detail with Black Point Compensation turned off but I don't know if that's an actual effect. As far as I know, Black Point Compensation is about pulling more detail out of the shadows and I don't understand why it would smooth out detail.
Aperture, Perceptual, Black Point Compensation: OFF
Aperture, Perceptual, Black Point Compensation ON

Aperture, Relative Colormetric, Black Point Compensation OFF
Aperture, Relative Colormetric, Black Point Compensation ON

I did some other tests, selecting different profiles, paper, etc, but the conclusion is that the Relative Colormetric is probably more accurate in colour rendition but Perceptual is more flattering for Portraits.

Colour Charts

I used a portrait in this test because I mainly use the camera for family portraits so getting good skin tone is my main concern. However, out of curiosity, I also looked at a colour chart. I took a colour chart that I printed from the internet, took a photo of it, imported into DPP and Aperture, applied minimal processing (adjusted white balance and exposure using a photo of the same set-up with a grey card), and printed it out. Ideally, you'd like to get an identical print, right?

What I got was some colours matching very well with the original in both DPP and Aperture (orange, brown), some colours out in both (pinks, greys) but out in different ways. For example, the deep blue in DPP was too pale, in Aperture it was richer than the original. I found that Perceptual Intent was closer to the original than Relative Colorimetric with the deep blue in RC having a purplish tint.

Original Colour Chart
Aperture 3, Relative Colorimeteric, BPC off

DPP, Faithful, Relative Colorimetric

Aperture 3, Perceptual

DPP, Faithful, Perceptual



It's interesting that I cannot reproduce the card. This is where colour spaces come into our already complicated picture.

Colour Spaces



Colour spaces describe the colours that can be described by a particular system. For example, if you're using the Adobe RGB colour space, you can add red, blue and green to create the colours shown within that triangle. sRGB limits you to less colours. The larger curve shows the colours that our eyes can distinguish. So the first point to note is that no colour space describes everything. The second is that colour spaces are theoretical. What you can actually show on a given screen or printer will depend on the device. Your colour space may describe the colour crimson but if your printer doesn't have red ink of the right shade, you're not going to be able to display it.

A gamut is a colour space in three dimensions. Gamuts become critical if you're editing images on screen and then expecting to see those colours when  you print that image. The gamut of your screen will not be the same as the gamut of your printer. In most cases, the screen can display more colours than the printer can print so the the software has to decide what to do with those additional colours. That is what the Intent I was referring to earlier is about. To borrow a wonderful analogy from Michael on Luminous Landscapes, it's as if someone give you a pillow and asked you to put it into a pillow case that is smaller. You can either squash the whole pillow into the pillow-case (Perceptual Intent - colours are shifted so that the relationships are kept but they're not quite exactly true-to-life) or you could cut the edges of the pillow so it matches the pillow-case (Relative Colorimetric  Intent - colours outside the printer's gamut are simply clipped to the nearest available colour.)

Perceptual vs Relative Colorimetric
That probably explains why I cannot seem to replicate the colour card exactly.

Adobe RGB vs Canon iP4500/MP610 GL2
This is illustrated in the graph above. The grey wireframe represents the Adobe RGB colour space. The colour block shows what the MP610 printer can do. There's a lot outside the printer's gamut and that means fudging it using Perceptual or Relative Colorimetric intent.

Except...

That is why I'm working with a colour card printed by the MP610. I know that my printer can display all these colours. So the problem is that it's not being asked to. Somewhere between taking the photo and printing, information is being lost.

The Canon EOS 5D mk II is set to Adobe RGB (however, when working in RAW, that is not a limiting factor as the colour space is only embedded in the JPEG, not the RAW file). The software interprets the RAW file in whatever the Working Color Space may be. For example, Aperture always works in Adobe RGB and I have DPP set to Adobe RGB.

Adobe RGB vs LCD Screen


My monitor may not be able to display all the colours in Adobe RGB but, again, I'm not colour-matching on screen so it doesn't matter.

The only colour translation is RAW to printer and, in both Aperture and DPP, I am printing to the same printer, using the official Canon ICC (basically, the translation for the gamut). So, why are they different?

Camera Settings

sRGB vs Adobe RGB

sRGB vs Canon iP4500/MP610 GL2
At this point, in desperation, I decided to see if switching the camera to shoot in sRGB would make any difference. It really shouldn't because I'm working with RAW images (not even RAW + JPEG).


DPP, sRGB, Relative Colorimetric, BPC off

Aperture 3, sRGB, Relative Colorimetric, BPC off
DPP, sRGB, Perceptual

Aperture 3, sRGB, Perceptual, BPC off
As before, Relative Colorimeter makes the deep blues purplish. However, Perceptual looks really good, especially the print from Aperture 3.

Original Colour Chart
The original chart is shown again and, despite the limitations of the scan, you can see that, while the pastels and greys still lack a certain brightness, for the first time, the other colours are very close.

At this point, we're talking about minor differences. Taken on its own, the Adobe RGB image printed using the same settings also looked close to the original:
Aperture 3, Perceptual
It's only when I compare the sRGB and the Adobe RGB together than I see they are not the same. Adobe RGB prints look darker somehow and so sRGB seems to be closer to the original.

If I were working with a JPEG from the camera, this would all make perfect sense. The Adobe RGB gamut is wider than sRGB. That doesn't mean that it can represent more colours, it means that it can represent a wider range of colours. Since the number of colours is the same, a wider range also means that the difference between individual colours is greater in Adobe RGB.

PetaPixel has a good explanation of this:


To illustrate this, I’ll use a simplified example. Suppose my color space consisted only of blue, and I could have a total of 3 possible colors. Lets say I chose to use the following “color space”:
Why You Should Probably Use sRGB srgbrange
Now, maybe I would like a wider range of colors to work with for one reason or another. Though I can’t increase the number of colors I can represent, I can increase the range by spreading the colors farther apart. The resulting color space might look something like this:
Why You Should Probably Use sRGB adobergbrange
Notice how I was able to capture a wider range (or gamut) of possible colors without increasing the number of colors. In both “color spaces”, I’m limited to 3 colors.
In the same way, Adobe RGB captures the same number of colors as sRGB but offers a wider range of colors by spreading the colors out more.

But RAW files don't have a colour profile. The software takes the RAW file and applies a profile when displaying it on the screen and again when printing. Both Aperture 3 and DPP are set to the same working space. The camera setting should make absolutely no difference.

At this point, I'm lost. I still don't understand why Aperture and DPP give different results in printing since, in theory, they are applying the same ICC to the same RAW file, which should give the same result. All I can guess is that it's something to do with the RAW conversion. But as to why the camera setting would affect the print-out of a RAW file, well that's just bizarre.

Conclusion

Shooting

sRGB or Adobe RGB

On the face of it, sRGB seems to have done better. I have no idea why but there it is.

The narrow gamut gave a better approximation to the original colour chart. However, that might not be true if the original image had any out-of-gamut colours (say, deep greens or turquoise). It probably depends on what exactly you're trying to capture but it may be that sRGB is sufficient for my needs. 

To be sure, I should review my existing images and see if any of the colours are out of the sRGB gamut. Unfortunately, Aperture 3 doesn't have this functionality. 

Printing

Relative Colorimetric vs Perceptual

Perceptual. 

Raw Conversion

Aperture 3 or DPP

The colour charts printed from Aperture actually looked slightly better than those from DPP. However, it was very slight. It might not be enough to justify a move from DPP to Aperture but, on the other hand, it does imply that the more complicated workflow of working in DPP & Aperture is not gaining much in terms of the final printed image. 

That said, I am concerned about the clipping on the red channel that I saw in the portraits. I've seen something similar in other images, when Aperture blows highlights but DPP doesn't.

I think I need to look a bit closer at the RAW conversion in Aperture and see if there are any other differences before I make the switch. 








No comments:

Post a Comment