Notice how the bright flowers are unaffected by the saturation boost, while the green background and the rest of the image gets more saturated.

I have also made it into an action which you can download at the end of the page.

Step 1:

Make a Gradient map adjustment layer with a gradient from black to white, to get a black and white image. Now set it to difference, select all (Ctrl+A) and copy merged (Ctrl+Shift+C). You can now delete the gradient map.

Step 2:

Enter the channels panel and make a new alpha channel, and paste (Ctrl+V) .

What you have now is a channel where the white parts represent the areas with high color saturation and the black parts the areas with low saturation. By using this as a mask, you can boost the saturation only in the areas where it’s needed the most and also avoid the artifacts you can get when just slapping on lots of saturation. But first you need to invert the mask. Also, you need to increase the contrast of the mask with levels to differentiate the effect more between the high and low saturated areas.

Step 3:

Then Ctrl-click the channel to make it into a selection, and make a Hue/Saturation layer.

You can now increase the saturation a whole lot more than usual before it looks bad, because it specifically targets those areas that are low on saturation.

Something else that is kind of nice is to turn the saturation on this masked layer down to a negative value. All the way down you will get a selective colored image where only the most saturated colors show, the rest being black and white. Turn it back up a bit and you can get a nice effect with a desaturated image where the most saturated parts are still very colorful.

Here is an action which does all the work for you DK_Vibrance.atn. It automatically maximizes the contrast of the alpha channel, and leaves you with a Hue/Saturation layer that you can tweak the saturation of until you are satisfied.

It works with PS CS and later, might work with earlier versions too although I have not tried it.

This is a walkthrough of how I edited this image in LAB color space.

The image I started with:

Step 1:

Convert to LAB color mode.

Step 2:

Make a curves adjustment layer. Adjust the A and B channel curves like below.

The colors will become very saturated in a way that looks much nicer than using hue/saturation in RGB mode.

Step 3:

Duplicate the background layer and apply Shadow/Highlight to the Lightness channel.

Select only the Lightness channel but keep all channels visible by clicking the eye icon next to the Lab channel.

Then apply the Shadow/Highlights adjustment by going to Image:Adjustments>Shadow/Highlight. My values are not cut in stone, you need to experiment a bit here. The image can look a bit flat but we will fix that in the next step.

You can adjust the opacity of the Shadow/Hightlights if you want to bring back some more shadows.

Step 4:

Edit the Curves layer with an S-curve on the Lightness channel to bring back contrast and depth.

Step 5:

Sharpen the image (preferably in the Lightness channel) and your done.

Digital camera sensors do not capture b&w images, they capture color images with three channels, each describing the lumniosity of each of the colors red, green and blue. Consequently, there is always a conversion taking place when you make a b&w photo. The three channels must be made into one grayscale channel. If you let the camera do that for you, you give up all control of how it is done, and how much of each channel is used in the mix.

If you do it in post processing on the other hand, you have full control. One way is using the channel mixer. It can be applied as an adjustment layer.

Check the Monochrome checkbox and you will get a grayscale image. The sliders control how much of each channel is used. The total percentage should be around 100%, but can vary pretty much from image to image.

To demonstrate the difference I’m going to convert this image of a bright yellow leaf you saw at the beginning of the post.

If we choose the three color channels separately it looks like this.

Red: Green: Blue:

As you can see there is quite a bit of difference. In the red one the leaf is very bright, in the green the leaf has a nice texture but blends in too much with the background. In the blue channel the leaf is very dark and the background very bright.

If we convert this image using Image>Mode>Grayscale we get an image that looks like this:

This is about what you could expect to get from an in-camera b&w image. As you can see there is not much contrast between the leaf and the background. The image becomes hard to read and it looks really bad.

How would we use the information in the r,g,b channels to improve this image? We would like to have the texture from the green and the contrast from the red. The blue channel is useful in this case, because we can use it to darken the distracting background by adding a negative amount of it in the mix, without affecting the leaf much.

So above is my version using the channel mixer. Click here to see the settings I used: Channel mixer settings. Adding a simple curvers layer to get some more contrast I get this:

That’s a whole lot better than the first conversion if you ask me, although there is a finer gray scale in the previous version. It depends on what effect you want to archieve.

Changing the settings you can get very different results. Here is another version where I subtracted a bit more of the blue channel to almost remove the background completely:

You might notice this last image is pretty noisy in the background. As a rule of thumb, the blue channel is the noisiest one while the green is the least noisy and most detailed and the red is the most contrasty.

There are many other ways to convert to b&w than just the channel mixer, the “black & white” option in Photoshop cs3 is one.

Thanks for reading!

So I started thinking. Isn’t it basically two polarizers screwed together?

Some basic theory on polarizers:
The basic polarizing filter is the linear one, it does nothing but polarize the light so that it is oscillating in one direction. If you put two linear polarizers over each other with the direction of polarisation perpendicular they will block out all the light. When you rotate one so that they are not perpendicular they will let through varying amounts of light. If they are parallell, you get no effect in the image.

But modern cameras have polarizers in front of their focus and metering sensors, so using a normal linear polarizer can disable those sensors when the light happens to be polarized in the wrong direction. Thats why there are circular polarizers.

Circular polarizers have a quarter wave retarder on the back of the filter, which scatters the light so that it is no longer polarized (or rater it’s circularly polarized). So you get the polarized effect with reduced reflections and deep blue skies etc. but with non-linearly polarized light entering the camera so that it doesn’t mess up the metering.

In this image I tried to show what I mean. The vertical component of the light passes through, and is scattered so that it still oscillates in both directions after passing the filter.

So first I figured that the Vari-ND must use two polarizers, one linear in front of a circular. That would first polarise the light in one direction, then the second would cut out some light depending on the rotation and lastly the quarter wave retarder would “un-linearly-polarize” the light before it enters the camera.

The problem with that is that the image would get a polarized look because of the linear filter in front. After some googling I realised you need two circular polarizers, with the front one put on backwards, to get rid of the polarization in the final image.

So I have this old 52mm circular polarizer that I’m not using much, and I placed an order on another today to try to make my own Vari-ND. Stay tuned!

Update:

I got the new circular polarizer today. It’s a cheap noname polarizer, it cost me about $15. I bought the cheapest one I could find.

After some quick tests it seemed to work OK, but there was a slight blue or yellow color cast when looking through the two filters at polarized light.

Here it is:

I used a small screwdriver to loosen the ring holding the glass in place. The new filter had bigger grooves and was much easier to remove the ring from so I decided to flip the glass in that one (plus it was cheaper so I’m not as careful with it).

Next, I removed the ring and took out the glass. After flipping the glass around I put it back in the filter and fastened the metal ring again.

…

—

So there it is, the assembled diy variable neutral density filter!

On top it the modified filter, and on the bottom the normal one. Time to try it out.

After some testing my worries were confirmed. Sometimes there was a blue cast, sometimes a yellow, and sometimes no cast. It seemed like it only affected polarized light. Though except from that, it’s light cutting capablities are as expected.

The light from my laptops LCD screen is polarized, as I found out when I held the CPL up in front of it.

So I used the screen for some testing.

If I held up a filter in front of the screen backwards, I got a cast. First it was all blue.

I rotated it 45 degrees and the cast dissapeared and it looked normal, after 90 degrees it got a strong yellow cast instead.

I tried the same with the other filter, and I could see a very slight yellow and blue cast when flipped, but far less than with the new one. So I would probably have got a much better result with the old one backwards in front.

—

I don’t know whats causing the cast, but when I took apart the filter I noticed the pieces of glass were pretty thick and that there was two seperate pieces that were sandwiched together, see closeup below. Maybe the fact that its made out of two separate pieces causes the casts. It feels really cheap (and it was).

The filter has its uses though. For black and white photography it works fine. The slowdown is from about 4 to 15 f-stops. Thats 11 f-stops range.

—

In my opinion, if you want flawless performance you should get the real thing. For a makeshift filter thats about 1/10 of the price though, it works pretty good as long as you stay away from the extreme edges of the range, or if you shoot black and white.

I have not yet been able to find a CPL that is cast-free when used backwards, but if anyone has, please put a comment here or mail me so I can try it out.

Filters tested:

• Soligor C-P.L gives a slight cast (the one I used in the back of my vari-nd filter)
• “Nikon CP-15 gives a color cast from blue to sepia”
• “Hoya HMC 58mm PL-CIR has mild yellow/blue cast”

References:

• http://www.fas.harvard.edu/~scdiroff/lds/LightOptics/CircularPolarization/CircularPolarization.html
• http://www.nikonians.org/dcforum/DCForumID8/16174.html
• http://www.meadowlark.com/applicationNotes/Basic Polarization Techniques and Devices.pdf

This image has lots of perspective and is also a bit rotated. By making a crop selection and clicking and draging outside the selection, it can be rotated and cropped in one action.

Press Enter or click the commit button to crop the image. You should get something like below.

The image is now rotated, but if you want you can also correct the perspective at the same time, with the Crop Tool. In this image the perspective looks nice, so normally we wouldn’t correct it, but just for the sake of demonstration we will. To correct the perspective you make the vertical lines, which are now converging upwards, parallell.

First make a crop selection in your image, then check the Perspective checkbox in the top toolbar. You can now move around the corners of the selection individually. You want to make the left and right edges of the selection line up with the lines in the image like below. By holding Shift when moving the handles you can make sure the horizontal edges stay horizontal.

Then use the middle handles of the selection to enlarge it.

Then make the crop and we are done. The image now has parallell vertical lines, something that a tilt-shift lens would accomplish.

The image is degraded more the less parallell the lines are, so it gives the best results if you try to keep the vertical lines as parallell as possible when shooting. A wide angle lens typically gives lines that converge more, so if possible use the longest lense you have when you know you will be correcting the perspective.

Daniel

Tilted horizons can often ruin the impression of a photo. Here is a useful tip when you need to quickly straighten the horizon in a photo using photoshop.

Select the Ruler Tool , and make a line following the horizon in your photo.

Now if you go to Image > Rotate Canvas > Arbitrary… the proper amount of rotation is already filled in.
Click OK.

Now all you have to do is crop the image and you have a photo with a perfectly straight horizon:

You can also use the Crop Tool and do both the rotating and cropping at the same time as below, but I find this method more precise on the rotation. See this post for info on the crop tool.

Daniel