Today, I want to show you a little example of picture processing carried out a few months ago. The original picture, on the left, was not taken by me; it is a self-portrait taken by the gifted photographer Keia Eskuetan (if you don’t know her work yet, you should go visit her gallery) and nicely contributed to my archive. Sometimes I like to play with someone else’s work as it allows me to try different things. The fact that the picture was taken by another person avoids the bias of including my own style in the picture, and allows me to obtain different and nice things using the constraints provided by the picture. In this case I’m really satisfied with how the picture developed and I’m using it as an example of one possible way to process a photograph.
In this case the starting material was a little bit complicated. The file I received was correctly exposed and nicely focused, but it happened to be saved in JPEG and was a little bit cropped vertically. There are some important factors to consider: Starting from a JPEG instead of a Raw file disallows the use of some high quality retouching that could be used for adjusting sharpness or light temperature among other parameters. This still can be done in Photoshop (or any other software), but using a lower quality starting point and less reliable techniques. Furthermore, having a cropped JPEG means that the picture has been edited and saved at least once (possibly twice). Every time we save a JPEG file, the compression process discards a little bit of information, worsening its quality. This is imperceptible to the eye, but when processing afterwards can be an issue. As a rule of thumb you should always work with a lossless format until the end, when you can save a copy on JPEG as a distribution file.
Step 1: After opening the picture up on the software the first thing to do is to prepare the file to work with it. This means, first of all, changing the color depth from 8 bits/channel to 16 bits/channel. JPEG is stored using the first value in order to save space. Increasing the depth to 16 allows us to use a much wider and richer set of colors and tones. Our final target will be a monochromatic JPEG, but even if it means less information, during the process we can benefit from the additional colors we can use (for example, in gradients). Also, in this step I decided the composition I wanted the final picture to have. First I chose an aspect ratio of 3:2 (or 2:3, depending orientation). I use this ratio as a signature of my work. After trying some cutouts I chose one that keeps the head, the camera and a part of the legs. Losing the shoes, an important part of the outfit and the original picture, is compensated by the strength gained by the final composition where they are not needed to explain the image. Finally, I decided to turn it 90 degrees, so it seems like our model is straight and floating in the air instead of crouching on the floor. This is the final composition and the idea that gives name to the picture: Weightless.
Finally, we need to duplicate the background layer and add it to a group, so afterwards, when we add some adjustment layers, we can work with all of them at the same time.
To avoid losing quality while working on the picture always work with a lossless format in 16 bits/channel. Change to 8 bits/channel and a loss format only to save a distributable copy at the end of the process.
Step 2: Although our final picture is a monochromatic picture, in order to achieve a good quality B&W conversion we need a good color image. The exposition of the picture is good, so there is no need to touch the levels. The color was pretty close to the right one, but it needed a little correction. In this case a duplicated layer of the image, with the color adjusted automatically and merged at 50% opacity did all the work.
After that, a “tone map” adjustment was applied in order to reduce micro contrasts in the higher contrasted parts of the image and increase contrast on the more flat areas. This tends to approach the overall image to the way the eye saw the scene, simulating a tiny HDR correction. This effect is much more subtle and less powerful than real HDR, but has the advantage that you only need one picture instead of at least two with different exposition (even on real HDR pictures tone mapping filters are used). This adjustment, anyways, is usually worth trying when photographing people, especially on interior or closed environments, as it tends to give the pictures a more natural lighting if the exposition is correct. It is to be noticed that this is always not the best option anyways.
My software doesn’t allow applying a tone map automatically, but it is an easy filter to build from scratch. Make two duplicates of your actual set-up (duplicate the working group, then combine the entire group to a layer and finally duplicate that layer again) and combine them inside a new group, inside the working one and on top of all the layers. Leave the lower layer as it is, and desaturate and invert the upper one. Change the opacity of the upper layer to 75% (we won’t change this anymore). We also set the opacity of the group to 90%, and the fusion mode to soft light. At this point we should see a lighter and very soft image. Now, on the upper layer, apply a Gaussian blur. The lowest recommendable value is 100 px, and good working values usually range from 100 to 200. The lower the value, the softer the image will be, the higher the blur, the more similar it will be to the original. We shall adjust the blur to a point where we soften some of the contrast on the clearer and darker areas but without softening the midtones. It usually takes a little bit of practice but it is easy to spot the right point. Finally, if we like the softness achieved but we want to make it more subtle we just have to decrease opacity of the group.
At this point we already have our composition and a good image for starting the real part of the process.
Step 3: We need to go from a color image to a black and white one. There are several ways to carry out this step, some better than others. The most commonly used by inexperienced people is simply by desaturating the image, which consists on maintaining the illumination value of the pixel while deleting all the chromatic information. This might seem like a good procedure but it presents a couple of problems: usually our eyes doesn’t sense brightness in the same way that the camera does, so a desaturation tend to offer flat greys in situations where a better contrast can be achieved. Besides, having three color channels to work with allow us to change illumination and contrast selectively on different areas of the picture, depending on the dominant color. Desaturation, on the other side, always chooses the same formula, without interpreting whether it is the best option for the situation or not.
In order to control the overall contrast of the picture, we have to create an adjustment layer on black and white mode. This layer allows us to convert the image to a black and white one selecting the absolute luminosity that the different colors of the picture will have. Do you want the red lips to appear dark while the Klein blue dress appears very bright? Just reduce intensity of the red color and increase the luminosity on the blue part.
This procedure gives us a lot of power over the picture, but with a great responsibility: Adjacent parts of the picture with different colors usually have the same brightness; when you increase the brightness of one and reduce the other sometimes “patches” appear, giving an ugly appearance. In order to solve this issue, it is recommendable to keep the values of adjacent parts of the adjustment controls relatively close (for example, if you set reds to 100%, a good value for yellows and magentas are between 50 and 150, but not usually -100 or 300). Also, for the same reason as before, adjusting too much the conversion tends to increase noise, as noise is uniformly distributed in brightness but randomly distributed in color. This means that when you separate luminosity on any two colors, you are separating the brightness of both kinds of noise, making it more noticeable.
It seems complicated, but with a little bit of practice it is very easy to achieve good results. In this case, the values used are the ones depicted in the image, and the final result of the adjustment is shown above.
Using B&W adjustment instead of desaturating allows more control on the conversion process. Using very different values for adjacent colors can lead to an increase of noise and patches appearing, so a careful control must be taken.
Step 4: In the last step we adjusted the overall contrast, but now it is time to adjust the contrast selectively on different parts of the image. To do this first we have to use our preferred method for selecting the area of interest, and after we create a curves adjustment layer. Playing with the curves allows modifying the brightness of the area in very creative and powerful ways. The intention of this post is not to show all the power curves have, but as an example, in order to increase contrast you set the middle point of the curve to its own value, and decrease the point between the black and the midpoint and raise the point between the midpoint and the white. The more accused the variation is, the higher the contrast. You can selectively increase the contrast by variating shadows more than lights, lights more than shadows or by setting the “midpoint” in any place where you consider that the neutral point should be.
In this case I increased the contrast of the legs and the skin (each part on its own layer and with own optimal values), raised the luminosity of the dress and decreased the luminosity of the hair. These modifications all together tend to increase the sense of depth of the body and create attitude by the rising of contrast and strengthening of shadows.
Step 5: The last important part is clarifying the background, removing shadows, details and increasing the isolation of the girl. The technique is the same as in point four, we select the background with our preferred method and increase luminosity in a controlled way until satisfied with the process. Removing the background eliminates distractions that can make the eyes wonder away from the main character, which is the place where we want them to stay. Also, a clearer background means and increase in the illumination and contrast perceived by the observer, without having to change anything in our main part of the picture, so it supposes an easy modification with a powerful increase in the overall attractiveness.
Step 6: Finally, we increase the sharpness of the picture. It is very important to leave this process always to the end and to apply it with the picture on its final size. If we want to have two copies of a different size, the best procedure is to duplicate the image, resize one (or both) to the desired size and after that increase the sharpness separately. Sharpness is very susceptible of the size and resizing after sharpening tends to increase artifacts and reduce the quality.
Many ways to increase sharpness are possible. One of the most used is the unsharp mask. In order to use it, you first need to duplicate the working group and combine all layers.
Instead of this method, I like to use a sharpening method based on a high pass filter. This method is better because it works similar to an adjustment layer, which can change opacity, be duplicated, combined or moved in order to fine-adjust the effect, instead of modifying a combined layer that cannot be modified afterwards.
First of all you need to duplicate the working group and combine. After that, a high pass filter is applied. The high pass filter removes all blacks and whites and leaves the greys, increasing contrast in border areas. A parameter (radius) can be adjusted. The higher the value is, the more noticeable the effect will be, but will be more prone to generate artifacts, like halos or Moire patterns. Moreover, the lower the size of the image is, the lower the value needs to be to achieve an actual sharpening. I usually work in the 0.5-1.5 px range for images lower than 6 MP and between 1-3 px for a 6-20 MP size. After applying the filter adjust opacity to 90% and fusion mode to soft light. In case you require more sharpening, instead of rising opacity to 100% or increasing the radius over 3 px, it is better to duplicate the layer, as the effect stacks. You can lower the opacity of the second layer if the effect is now too noticeable.
Always perform sharpening on the last step of the process, and always at the final size of the image. Some photographers prefer to save a picture without sharpening and, when they need a copy for any purpose, they just make a copy, resize adequately and apply sharpening on that copy.
After all this process we get our final picture, which can be seen on the right. This procedure represents one of the possible examples of how a picture, taken with a concrete idea in mind, can be changed dramatically just by some simple processing effects and a little bit of imagination.
As a final remark, remember that JPEG doesn’t allow 16 bits/channel, so if you want to save it in this format, first you need to combine all layers and, after that, change to 8 bits/channel mode and save the file. Doing it this way allows us to do all the modifications with the highest amount of colors and possible tones, and just reducing the depth of color for the saving step.
What do you think of the picture? Do you like it more before or after? Is there a retouching you did that you are especially fond of? Share your impressions on the comments and give any idea you would like to share.