Comparing B&W conversion methods

Although I have already blogged about it, converting images to Black & White is a bit like the recipe of the Swiss Fondue: everyone has his own touch to it and claims his way to be “the genuine one”. So I decided to revisit the subject (B&W converting that is, not the fondue) and try several conversion methods and see what results are offered. The methods are:

  1. Simple desaturation with RawStudio
  2. Use a channel mixer
  3. Use the Luminance curve
  4. Use the special Jpeg2BW tool by Thomas Baruchel

The picture I will use as an example is that one, taken on the side of the Lake of Lucerne.

By the way, speaking of Fondue and Lake of Lucerne is timely, since August 1st is the Swiss National Day🙂

Foreword

My goal is to see how much difference there is between several conversion methods in one real life case.

Practically, I tweaked my file in RawStudio to create the base image (above). I then used this same image to generate the different B&W versions. So what we see here is really the effect of different conversion methods.

And instead of just presenting each result on the way, I have decided to do a little challenge (à la Meetthegimp.org): the b/w versions are available at the bottom of this post without mention of the method used. I would be interested to read your comments on how you like the different outputs and why.

Desaturation & Curves in RawStudio

I said myself that you shouldn’t do that… but I do it more often than not. It is so easy – just put the saturation to 0 and you can get a decent B&W picture in no time; the curve tool is so powerful and RAW offers such a latitude of treatment.

Will that make a good final output, however?

Channel Mixer

This is the “proper way” to do B&W conversion. You basically mix the Red Green and Blue components of the image – usually with alot of green, a bit of red and almost no blue. Although red is the most “spectacular” channel, green has less noise (since there are twice more green photosites than blue and red) so using a good chunk of it is a safe bet.

This is the method I described previously.

Luminance Channel

This is another technique presented as offering very good results. Luminance offers an excellent basis for a B&W image; just decompose your image to LSV and use the L channel – unfortunately, this had to be done in 8 bits since I didn’t find a way to do it in Cinepaint.

Jpeg2BW by Thomas Baruchel

In a comment, Thomas Baruchel said that he had developed his own B&W conversion routine that he as put as a command line tool on sourceforge – it basically takes a JPG file as income and spits out a B&W “optimized” image. You can read the details of his conversion method in French on his blog (part 1, part 2) or in English with the Jpeg2BW program itself.

Jpeg2BW is straightforward to compile, just use Guile 1.8 (Ubuntu Hardy’s verison is 1.6):

gcc -o jpeg2bw jpeg2bw.c -ljpeg -lguile -lm

Usage is (info on the PNM image format):

./jpeg2bw color_image.jpg > bw_image.pnm

The results

I then scaled each image down for the web, added a bit of local contrast and unsharp mask (same amount for each image, of course). And here are the results, in no particular order:

A.

B.

C.

D.

Which one do you like best, and why? (I’ll put the correspondance method – result in a later post)

21 Responses to Comparing B&W conversion methods

  1. Georges says:

    I prefer A, it has the smoothest structure, which goes best with the chosen subject.
    Then come B and D (I do not see any difference between these two). Those have more contrast than A but also look quite good.
    I do not really like C. It looks very “edgy”, to much contrast and unnatural.
    So much for my opinion… and happy 1. of August, I hope you have better weather in Geneva than we here in Zürich.

  2. PVG says:

    Between B and D I don’t see any real differences. I like more C, but it could be better with less contrast.

    So:
    1) C
    2) B e D
    3) A

  3. BDion says:

    As others have noticed B & D look very similar. My choice,

    1- B or D
    2 – A
    3 – C

    A looks a little flat

    C looks too light and makes the foreground hill on the right look like a big black blob.

    As a side question, are you now using the 1.0 version of RawStudio?

    Thanks
    Bob

  4. fordybsd says:

    A or D, i like the darker sky in what was the blue part of the sky.

  5. Thank you for speaking of my program. For extracting the luminance channel in 16bpp, you can do it very easily with ImageMagick ; by the way you can extract the L channel from HSL or the V channel from HSV or the L channel from LAB, the Y from XYZ, or the Y from YCbCr. If you have a JPEG, the Y channel from YCbCr is already available (in 8bpp but this is the most accurate you can get since you just have to ‘dump’ the channel from the JPEG file with no further computation); just use the jpegtran tool (from the libjpeg package) with the relevant option (-grayscale).
    I am not home for the moment but I think the extraction of the luminance channel with ImageMagick is performed with something like (not tested):
    convert -depth 16 my_input_image -colorspace HSL -channel B -separate output.tiff
    (remember that B is merely a shorhand, taken from the word ‘RGB’, for the third channel; if you want the L channel from LAB, which is the first one, you will have to use the shorthand R).
    You can also extract all these various luminance channels, computed in a highly accurate double precision format, with the jpeg2bw tool itself (by using the relevant flags of course, see the manual).

  6. Jon Leighton says:

    1. B and D
    2. C
    3. A

  7. Rolf says:

    I like D most, can’t decide a ranking between the others.

    I guess that D was made with the channel mixer – but I am not sure.

    My I spam your Blog with a pointer to my Monochrome Conversion series at Meet the Gimp?

    http://meetthegimp.org/episode-032-fade-to-gray-monochrome-conversion-1/
    http://meetthegimp.org/episode-033-the-channel-mixer-monochrome-conversion-2/
    http://meetthegimp.org/episode-034-full-control-monochrome-conversion-3/

    The last one shows a different way of conversion with layers and layer masks. Something for the Ansel Adams in you.😉 32 and 33 are just the basics.

  8. Neil McShane says:

    Joel, you answered your own question. Many different ways with different results.
    For me I pick none of the above.
    Why, well I would combine the foreground in C (as you have detail in the shadows on the sides and good contrast on the mountains) with the sky in A ( and a bit more burning in the sky to make it even darker.)
    As you know from looking at the images form my website I do like a dark sky.

  9. Robenroute says:

    1. B
    2. D
    3. A
    4. C

    The contrast in picture C is off (compared to the original base photo). A is too soft and B has just the right amount of detail in the dark areas, where D trails slightly behind.

  10. Benedict says:

    I like A the best. C is the worst. It looks too washed out.

  11. henrique says:

    C.

    I like the sharper contrast, because it makes the BW more “useful”.

  12. Rolf says:

    The sky from a, the hills from c and the water from d.

    A bit more work…..

  13. Hi all:
    I prefer D, very similar to B but it has a little bit more contrast (I can apreciate it in middle’s mountains.
    I think that the correspondence is:
    1. D
    2. A
    3. C
    4. B

  14. Xavier says:

    Hi,

    I would go for :

    D (well balanced)
    B (less contrasted in the grassy mountains)

    then

    C (too contrasted)
    A (too flat)

  15. Martin Ling says:

    Recent CVS versions of UFRaw support methods A-C natively plus a couple of others, thanks to a patch from Bruce Guenter in March.

    Personally I would go for somewhere between your B and C images. C has great contrast but loses some detail in the darkest areas

  16. Andrew A. Gill says:

    I prefer D. B is a close second but the water is just a bit too dark in B. I can’t abide the sky from C, as it’s too washed out. The hills in A are a little light in contrast.

    This is all aesthetically, of course. If I looked at the technical specifications of each image, the sharpness of the features and so on, I might come to a different conclusion. D is prettiest.

  17. Colin M says:

    Having not read the answers yet, I’d agree with PVG and say:

    1) C – better, brighter mid-tones
    2) B or D – virtually the same, decent mid-tones and contrast
    3) A – lowest contrast image

  18. dabnotu says:

    thanks for documenting this great Gimp feature
    here’s my first edit using it:
    JK x _3783023884_149ed21441_o _GIMP
    peace!

  19. super real says:

    Small Business Loan Rates Businesses needing capital examine small enterprise loan rates to view when it
    is smart to consider on debt super real once
    your lenders respond to your request, you may probably schedule a
    meeting with the crooks to work out a brand new offer regards to your home loan.

  20. Alice says:

    fantastic issues altogether, you just received a logo new
    reader. What might you recommend in regards to your put up that you made some days ago?
    Any certain?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: