Erwin Puts: Raw converter comparison LR3, Capture one 5.....

[tgm]

Advertisement (gone after registration)

Erwin Puts has posted in interesting comparison of most recent Raw converters including Lightroom 3 and capture one 5:

 

M9, part 8B1

 

 

Thomas

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[gwelland]

Well II guess it's interesting at a superficial level but I left thinking where's the beef ... the images are so small that they don't really show much other than the colour aberration.

[jklotz]

I found it interesting that the only one that did not exhibit color aberrations was SharpRawPro, which, according to their website, has no support for an M9, or even an M8....

[novice9]

Capture 1, while showing color artefacts, does seem to resolve better than the others.

[tgm]

Capture 1, while showing color artefacts, does seem to resolve better than the others.

 

yes, I agree, and I was surprised because there are many very positive reports on the improved image quality of lightroom 3. Then, I downloaded the demo version of capture one 5 and did some comparisons with lightroom 2.5 and lightroom 3 beta. The DNG file was taken with the summicron 2 /28 at f stop 5.6, 400 ASA. Here my personal conclusion:

 

1. LR3 beta is clearly superior to LR 2.5, the chroma noise is much less

 

2. CO 5 has still a slight edge over LR3, the image looks less grainy. All in all the image looks more natural and smooth although it shows the same amount of details as the LR 3 image. Whether this can be seen even in a large print, I don't know.

 

3. The most obvious difference between CO and LR are shadows, in LR images they are somewhat dark.

 

I have not made a final decision if I will switch to Capture one, I will make some prints and see if the differences are visible.

 

Thomas

[sanyasi]

etails as the LR 3 image. Whether this can be seen even in a large print, I don't know.

 

3. The most obvious difference between CO and LR are shadows, in LR images they are somewhat dark.

 

Thomas

 

I would be curious what your final experience is. You say shadows are somewhat dark in LR, but does that take into account the adjustments you can make? I don't know, but maybe the starting points are different in each program and it is up to you to make the adjustments. If you can get to the same place in each program, does it matter, particularly if both programs make you do some work in differing aspects of the photos?

 

Jack Siegel

[sandymc]

Advertisement (gone after registration)

I would be curious what your final experience is. You say shadows are somewhat dark in LR, but does that take into account the adjustments you can make? I don't know, but maybe the starting points are different in each program and it is up to you to make the adjustments. If you can get to the same place in each program, does it matter, particularly if both programs make you do some work in differing aspects of the photos?

 

Jack Siegel

 

The default tone tone curves for LR and C1 are different, which leads to different shadow versus highlight rendering. I wrote about this a while ago here: ChromaSoft: Lightroom, Aperture and Capture One Mini-Review Part 2

 

Sandy

[sanyasi]

The default tone tone curves for LR and C1 are different, which leads to different shadow versus highlight rendering. I wrote about this a while ago here: ChromaSoft: Lightroom, Aperture and Capture One Mini-Review Part 2

 

Sandy

 

 

Thanks. Your blog post is very interesting and useful. Next editing session I think I will start from ground zero and build the picture.

 

Jack Siegel

[tgm]

The default tone tone curves for LR and C1 are different, which leads to different shadow versus highlight rendering. I wrote about this a while ago here: ChromaSoft: Lightroom, Aperture and Capture One Mini-Review Part 2

 

Sandy

 

 

Sandy,

 

thanks for the comment, now I learned why the shadows in C1 look different from other programs.

 

Erwin puts comparison was motivating me to process some M9 dng files with LR 3 beta and C1 5. Here my personal conclusion.

 

1. The difference in image quality, detail and sharpness is very small. Small means, that in side be side comparison of 60x40 cm prints on glossy paper they are difficult to see.

 

2. C1 5 has a slide advantage in purple fringes in the corners. Fringes are better removed than LR 3, you can remove them also in LR 3 using the menue lens corrections. But I could get only reasonable results if I use the option 'defringe all edges' but than the whole image looks somewhat muddy. For landscape photography, taking pictures of trees especially in winter time, C1 gives slightly better results ( again, you only see the difference in large prints).

 

3. LR 3 and C1 5 have a slightly different characteristics of preserving details. C1 resolves horizontal and vertical lines slightly better the diagonal (thats why it has higher resolution, see Erwing Puts comparison) but LR 3 has a slight edge on diagonals.

 

4. C1 5 creates slightly less chroma artefacts than LR 3.

 

5. C1 5 produces a lot of almost isolated, single white and also some black pixels, which give the image a special appearance.

(see also the comparison by HankScorpio

http://www.techdystopia.com/photos/20091031-AR2H1113-LR3b.jpg)

they are to some extend artificial, but make the picture vivid. In that respect LR 3 produces more natural images.

 

In summary, the differences are really small, but exit. I think it is a matter of taste.

 

All in all I will probably stay with LR 3 ( and 2.5). One reason, I had several times the problem that it took an enormous amount of time ( several minutes) for C1 five to produce an output, while LR is quite fast. Sometime C1 got stuck, is that normal?

 

Thomas

[UliWer]

...

 

1. The difference in image quality, detail and sharpness is very small. Small means, that in side be side comparison of 60x40 cm prints on glossy paper they are difficult to see.

 

...

 

This leads me to a- perhaps very stupid - question: How do you print from C1?

 

I always process in C1 to 16bit-Tiff and import this in LR for printing. Even if I don't change anything in LR and use the printer's management I think that there should be some influence of LR on the result.

 

So when I should compare C1 with LR from prints made by LR I think I am just comparing two different LR versions.

[sandymc]

Sandy,

 

thanks for the comment, now I learned why the shadows in C1 look different from other programs.

 

Erwin puts comparison was motivating me to process some M9 dng files with LR 3 beta and C1 5. Here my personal conclusion.

 

1. The difference in image quality, detail and sharpness is very small. Small means, that in side be side comparison of 60x40 cm prints on glossy paper they are difficult to see.

 

2. C1 5 has a slide advantage in purple fringes in the corners. Fringes are better removed than LR 3, you can remove them also in LR 3 using the menue lens corrections. But I could get only reasonable results if I use the option 'defringe all edges' but than the whole image looks somewhat muddy. For landscape photography, taking pictures of trees especially in winter time, C1 gives slightly better results ( again, you only see the difference in large prints).

 

3. LR 3 and C1 5 have a slightly different characteristics of preserving details. C1 resolves horizontal and vertical lines slightly better the diagonal (thats why it has higher resolution, see Erwing Puts comparison) but LR 3 has a slight edge on diagonals.

 

4. C1 5 creates slightly less chroma artefacts than LR 3.

 

5. C1 5 produces a lot of almost isolated, single white and also some black pixels, which give the image a special appearance.

(see also the comparison by HankScorpio

http://www.techdystopia.com/photos/20091031-AR2H1113-LR3b.jpg)

they are to some extend artificial, but make the picture vivid. In that respect LR 3 produces more natural images.

 

In summary, the differences are really small, but exit. I think it is a matter of taste.

 

All in all I will probably stay with LR 3 ( and 2.5). One reason, I had several times the problem that it took an enormous amount of time ( several minutes) for C1 five to produce an output, while LR is quite fast. Sometime C1 got stuck, is that normal?

 

Thomas

 

C1 getting stuck isn't normal - I'd suggest taking that up with Phase One - they're generally very responsive.

 

As regards detail in C1 versus LR, what Irwin doesn't discuss in his article (or perhaps isn't aware of) is that in demosaicing algorithms for raw converters there is a tradeoff between extracting detail and artifacts. Specifically, the tradeoff comes about as a result of how the converter treats the two green channels. If it treats them as being the "same", then it can extract a lot of detail, but risks artifacts if the two green channels diverge e.g., they have an offset. This is how you get maze patterns - e.g, if you try to convert a S2 file with the current version of C1, which isn't yet tuned for the S2. This is also why DNG files have the BayerGreenSplit parameter - it tells the raw converter how close together (or not) the two green channels are.

 

So you really have to look at artifacts and detail extraction ability together. My experience is that usually C1 favors detail extraction at the expense of artifacts, and LR tends to avoid artifacts even if it means loosing detail.

 

Sandy

[tgm]

C1 getting stuck isn't normal - I'd suggest taking that up with Phase One - they're generally very responsive.

 

As regards detail in C1 versus LR, what Irwin doesn't discuss in his article (or perhaps isn't aware of) is that in demosaicing algorithms for raw converters there is a tradeoff between extracting detail and artifacts. Specifically, the tradeoff comes about as a result of how the converter treats the two green channels. If it treats them as being the "same", then it can extract a lot of detail, but risks artifacts if the two green channels diverge e.g., they have an offset. This is how you get maze patterns - e.g, if you try to convert a S2 file with the current version of C1, which isn't yet tuned for the S2. This is also why DNG files have the BayerGreenSplit parameter - it tells the raw converter how close together (or not) the two green channels are.

 

So you really have to look at artifacts and detail extraction ability together. My experience is that usually C1 favors detail extraction at the expense of artifacts, and LR tends to avoid artifacts even if it means loosing detail.

 

Sandy

 

Sandy,

 

thanks for your comments, at the moment I am just using a 30 days test version of C1 5 so I should not complain and contact Phase one.

 

Then, what do you mean with two green channels? When you compare C1 and LR are talking about the most recent versions 5 and 3 beta, respectively?

 

Thomas

[sandymc]

Sandy,

 

thanks for your comments, at the moment I am just using a 30 days test version of C1 5 so I should not complain and contact Phase one.

 

Then, what do you mean with two green channels? When you compare C1 and LR are talking about the most recent versions 5 and 3 beta, respectively?

 

Thomas

 

The Bayer matrix in the sensor has two green channels - the matrix is

 

RG

GB

 

My experience is that the detail versus artifact tradeoff has been the same since the earliest versions of C1 V4 and LR V2

 

Sandy