571 - Them There Eyes II



This is an image that I shot on Tuesday. I couldn’t post it that day, because my ISP had severe DNS problems. I could resolve most websites, but www.blogger.com was not one of them. Bad luck. The next two days I had no time at all, now it’s Friday afternoon, I sit on the train again and try to catch up.

While post-processing this image of an old Volvo, I thought that the method is very simple, takes hardly any time and is so versatile, that I should probably make another tutorial. Here we go:

This time we look at the incredible power of a frequently underused feature - blending modes. What is a blending mode anyway? Well, that’s basically the method Photoshop uses when displaying a layer that is on top of other layers. The most frequently used blending mode (and the default) is “Normal”. Basically that means that the top layer hides the layers below. Pretty as you would expect it. Think of it as a stack of playing cards. You can see the top card, but nothing below.

Even in “Normal” mode we can make some interesting things. We can attach a mask to the top layer, and by painting on the mask with black, we can hide the layer partially (mask it). It’s a bit like cutting away parts of the top card in the stack. We can also lower the opacity of the top layer from 100% to, say, 50%. Now you can see through the top layer like through a colored plastic foil.

Both of these techniques, masking and opacity, will be used, but we will use it on layers of more exotic blending modes.

Let’s first begin with a look at the original Volvo as the camera has recorded it. I always shoot “Large RAW + JPEG Fine”, which on my D300 means to use the full 12 megapixel, record the sensor data as a RAW file and to additionally produce a JPEG file of the same size and the best possible quality.

This image, the JPEG from the camera, is obviously meant to be about the headlight. I have used my Sigma 10-20 at f4, went very near and focused on the glass. With a lens this wide and with a maximum aperture of f4 (note: maximum aperture = minimum f-number), you have always a big depth of field, but when you focus near enough, you’ll still get some background blur. That’s what I was after.

As regards exposure, the camera has done what cameras set to matrix metering tend to do. The exposure is pretty leveled out. There is clearly detail almost everywhere in the car, only the reflexes in the headlight are partially burnt out, but that still looks pretty good to me. The house in the background is perfectly exposed but ugly, and although the sky is much too light, it seems to hold detail as well.

I want this to be about the light. As the image originally was, the light was in an awkward position, neither centered nor on a third. I wouldn’t want to center it anyway, so let’s put it on a third. The Rule of Thirds is no hard rule at all, but here it does well. In Camera RAW I crop in from the left and a little from below, a tiny bit from the right, and at the same time I make the image boring and flat. I do this by using a linear tone curve (medium contrast is the default) and letting the automatics set exposure, contrast, etc.

I don’t always crop in Camera RAW. Normally I do it in Photoshop, and sometimes even at the very end of processing. In Photoshop I also choose the option to hide the cropped area instead of deleting it. Then I can always go back with “Image / Reveal All”, and that without reverting the other steps made in between. Here cropping in Camera RAW is OK, as I exactly know what I want.

For the next steps I want to have another layer, basically the same, but with even less overall contrast and instead much increased local contrast. This is the layer that I want to use for blending. In order to get such a thing, I duplicate the background layer and use the PhotoLift plugin on it. See “492 - Roughly About Sundown” for more about that. Alternatively you could also use a curves layer to lower the contrast, and then unsharp mask with a high radius and a low amount (well, to get this effect you’d need more of a medium amount). This is not as convenient as PhotoLift and takes some experimenting, but it works quite well. See how we get detail in the sky now. The look of this layer is almost like many HDR images are, unnatural and comic-like. As it is, this layer is still in “Normal” blending mode.

Next we set the blending mode (that’s at the top of the “Layers” palette, left of “Opacity”) to “Multiply”. Eeek! That’s much too dark! On the other hand, the sky is nice, the ugly house is mercifully lost in the shadows, and maybe the darkness would do nicely as a vignette.

Let’s add a mask now, and then let’s paint with a big, soft brush and the color black on the mask. Where it’s black, the layer with the mask is hidden. Let’s do that on and around the headlight. Ahhh! Much better. The problem is only, that what we have revealed, is still the boringly flat original background.

What do we need now? Basically we want our contrast back, and along the way we want some more colors as well. We don’t want it everywhere, we only want it on the headlight, or in other words, we want it where we have painted with black on the mask. I simply duplicate the top layer, change the blending mode to “Soft Light”, and then invert the mask. Voilà! A little sharpening with an edge mask, and that’s the Image of the Day. Here is a shot of the layers palette.

What have we done? We have set a strong focus in the image. This is now really about the headlight, nothing else.

Of course the same result could have been reached in a number of ways. There is always more than one way to do things in Photoshop, but I think two layers, that’s not too shabby. The point is, that it really pays to know about blending modes and what you can do with them.

This is the image that inspired me to write about blending modes, but compared to the original image, the effects are still subtle. Let’s look at something really dramatically bad, and let’s try to chang e it into something usable.

Today I’ve asked my friend Erich to sit for a really bad portrait. I wanted something terribly lit, an image with a light background (a window), the face looking into the room, being fully in shadow. This is a worst case scenario, something that I would normally avoid under all circumstances, and if I couldn’t avoid it, I would use a flash. Still, sometimes such an image is all that we get. It has either been taken by someone else who didn’t care, or we had the choice to take it or get no image at all. The first image is straight from the camera. We have extremely harsh contrasts, the background is partially gone and we still don’t see details in the face.

The first step is again to convert it in Camera RAW into something flat. The real reason behind this is to incorporate all detail that we can get. The result is even less attractive. Now let’s do some blending.

We begin with “Multiply” again. But, wait, what do we blend? For the last image we have used a pixel bearing layer with increased local contrast, but this is not always necessary. You can blend any layer, even adjustment layers. Thus we add a curves adjustment layer, don’t bend the curve at all, and only set the blending mode to “Multiply”. This has the same effect as duplicating the background and setting the result to “Multiply”, only the curves layer takes much less space in the resulting file. But this is not only more efficient, we could even manipulate the curve to fine-tune the effect. No need to do it here, but it’s good to know that we can. The effect on the background is OK, but of course we want to paint in the mask to reveal the face. This is what the image to the right shows.

Let’s add another curves layer to lighten up the face. I duplicate the “Multiply” layer, change the blending mode to “Screen” (which strongly lightens up) and again invert the mask. Now that’s dramatic! For the first time we see the face.

That’s positively the right direction, but I want more light. One way would be to duplicate the screen layer, but doing so still does not give enough light, and even worse, the contrast in the face is deteriorating. Let’s try another blending mode.

Basically there are three groups of blending modes that work well in such situations. One group darkens the image. “Multiply” is the most frequently used mode of them, “Color Burn” and “Linear Burn” are also useful. A white layer in one of these modes is neutral and does not change the image.

The second group lightens the image. We have already seen “Screen”, “Color Dodge” and “Linear Dodge” are others. A black layer in one of these modes is neutral and does not change the image.

Finally there are modes that increase contrast. Light portions of the upper layer lighten the image, dark portions darken it. “Soft Light”, “Overlay” and “Hard Light” are the most useful modes in this kind of post-processing. A mid-gray layer in one of these modes is neutral.

What we need here is first some more light, and trying the lightening group shows that “Linear Dodge” does quite well, although we need to dial back opacity to 80%. The first attempt at a mask was a copy of the mask for the screen layer, but then I decided to use a strongly blurred version of that. Furthermore I have painted in the mask to tone down some highlights that would otherwise have burned out. “Linear Dodge” preserves more contrast than “Screen”, but it also tends to be aggressive to extreme highlights, so be careful.

The next step is to increase contrast. We don’t need much, but some contrast we do need. The most gentle mode to increase contrast is “Soft Light”. “Overlay” and “Hard Light” would be next, but for this particular case, “Soft Light” at an opacity of 50% is OK.

Originally the face was in complete shadow and we had no clue what a correct white balance would be. Now though we see that the face is too yellow. The camera was on automatic white balance, and in that insane mix of background daylight and muted interior neon light it actually did quite well. Still, it’s too yellow and we’ll need to correct that.

There are many ways to correct color, and while I have extensively used Lab color mode last year, my current tool is the “Photo Filter” adjustment layer. We need some cooling here, and the cooling filter of choice for this image is “Cooling (LBB)” at the default strength of 25% and an opacity of 30%. Of the three cooling filters, LBB is the one that has a slightly reddish cast, and that looks good here.

Now that colors and tones are about right, it is a good time to clone out some blemishes of the skin. Remember, this is not about altering the image, it is about removing distractions that are not part of the personality of your model. Everybody has some red spots at times, but nobody considers those spots essential for recognizing the person. They are alway in different positions, it’s only the photographic image that would lock them in place. By removing them, we only reveal the archetype the sits below. Good riddance.

We could stop here, but a little beauty blur is always nice in a portrait, and even more so when the image quality is already stressed by an attempt to pull detail out of deep shadows. I call this my “neutral blur”, and I have an action for that. Basically it goes like this:

Select the whole image, “Copy merged” and paste into a new layer. Duplicate this layer. Set the first one to “Screen” mode and Gaussian blur it with a radius of 30 pixels. Let Opacity at 100%. Then set the other one to “Multiply”, blur it with a radius of 5 and set the opacity to 60%. Group the two layers and set the opacity of the group to what looks good. Here I have used 70%.

For women we would probably leave it at that, but Erich is a man and here we want a tad more definition. I could have used a PhotoLift layer, but instead I “Copy Merged” again and use unsharp mask with an amount of 60 and a radius of 60 pixels on the result. An opacity of 50% is ideal in this case.

Impressive? Certainly. It’s not that I did this in zero time, not at all. Especially the mask of the “Linear Light” layer took me some time, but I guess the result clearly recommends having a look at blending modes.

One note though: Don’t expect such extreme manipulations to work with JPEGs taken with a point-and-shoot camera. Photoshop can’t do wonders. Noise and JPEG artifacts will frequently restrict how far you can go. For maximum malleability you need RAW files and a DSLR.

The Song of the Day is again “Them There Eyes”, but this time it’s not Louis Armstrong, this time it’s Anita O’Day and her 1957 collaboration with the Oscar Peterson Trio “Anita Sings the Most”.


There are 1 comments

Ted Byrne   (2008-05-11)

And here is your second collection (I am reading backwards today) just as you need to collect your bicycles... you also need a vehicular collection. You find parts of these things and anthropomorphize them. Now that you've pictured enough trees it is time to see the forrest. What your tiny scavengings add up to when they are brought together. No one better finds the lives who breathed themselves into these metal sheets. Show us what's emerging...

💬 Reply 💬