A New Year's "Resolution": Sharpness EQ

Background

Once in a while when we find or create something truly unique, the idea gets left behind as we move on to new things. Ever bake something that was so good that you ate it twice a week for a month and then just moved on to something else out of sheer boredom? Ever come back to it a year later and remember how good it really is and felt like you discovered it all over again? This article falls into that category where we revisit an old but very useful idea. Let's take a look at sharpness variance in digital photos and ways to correct sharpness variances to bring out more presence or 3D effect in photos.

The problem

The vast majority of cameras on the market use CFA's (color filter arrays) to capture only one color at each pixel location. The Bayer CFA above is by far the most common sensor type. Notice that only one color (red, green, or blue) is captured at each pixel location on the sensor. Sophisticated algorithms must be used to "predict" the missing two colors before you get to the final full color image that you see from your camera or raw conversion software. To complicate matters, there are twice as many green pixels as red or blue, in part, in order to mimic the human eye and its greater sensitivity to green compared to red/blue.

If you take a picture of a subject with very little saturated color like a B/W resolution chart, snow scene, the moon, or other objects without saturated colors, it is easy to predict the missing colors because all three primaries (red, green, and blue) will have about the same brightness. In these cases, the missing green and blue values will be about the same as the red brightness captured by a red pixel, red brightness at a green pixel will be about the same as the capture green value, etc.. Once you start photographing subjects with more vibrant colors such as fall foliage, colorful Halloween costumes, or the worst case scenario: a red rose, the amount of detail captured by the camera is significantly reduced. As an example, consider the red rose. A red rose of a particular shade will only excite the red pixel locations on the sensor, leaving very little (usable) information at the green and blue photosites (pixels). For the red rose, your camera's resolution just dropped to near 1/4 of its total resolution due to the fact that the green/blue pixels on the sensor are contributing very little information. In cases like this, the problem actually becomes visible in photos! Your red rose may look a little soft or out of focus compared to the green leaves or brown parts of the stem that are in the same focal plane because leaving you to wonder if perhaps your camera didn't focus on the red flower as it should have.

If you train yourself to pick up the problem, it is quite noticeable! A bright blue sweater in one particular photo may look a little out of focus compared to a gray sweater right next to it, you may find it difficult to get a truly sharp photo of a blue flower while the green leaves around the flower look sharp, and so on. This sharpness discrepancy for different colors can alter the relationship between sharpness and depth of field and can take away some of the 3D effect or "presence" that is seen on cameras that capture full color (all three colors at each pixel) like the Sigma SD9, SD10, and SD14. If you keep up with the reviews or visit online forums, you will likely hear a lot of buzz about how full color capture cameras like the SD9, SD10, and SD14 create photos with more 3D effect than other cameras. The reason for that is in large part due to the fact that full color capture cameras do not suffer from sharpness discrepancies and capture all colors with the same amount of detail. This leads to a much greater correlation between depth of field and focus which is what adds presence or 3D feel to photos.

The remedy

Fortunately, some years ago I found that you can take a (preferably unsharpened) photo and apply a special adaptive sharpening algorithm to effectively reverse the effect of color sharpness discrepancies. The image sensor in your camera cannot capture all colors with the same detail, making certain colors (like saturated red and blue) look considerably softer than other colors such as gray or even green. The fix is to apply sharpening in such a way that it sharpens saturated reds and blues the most, greens to a lesser extent but still more than grays, and so on. While sharpening can't truly add information that has been lost to single color capture sensors, the adaptive sharpening technique can produce a more visibly pleasing result so that bright red detail doesn't look considerably softer than gray/white, green detail doesn't look twice as sharp as blue, and so forth.

I created an algorithm that effectively reverses sharpness discrepancies and called it the "sharpness equalizer", adding it to the repertoire of image enhancements in my own Qimage's batch filtering tool. Simply select your USM (unsharp mask) and slide the equalizer slider to the right to bias the sharpening algorithm to compensate for sensor sharpness discrepancies. Using values like 2 for the radius, 150 for the strength, and the equalizer slider all the way to the right (to try to compensate completely for sensor sharpness discrepancies) increases the 3D feel of images and improves overall clarity of photos. I made my algorithm available to Uwe Steinmueller who also created a PhotoShop plugin that does the same type of adaptive sharpening. See my earlier article on the Outback Photo web site for details on the plugin.

Since I have more than one dSLR camera and I'm always comparing the latest models to my full color capture SD14 for sharpness and 3D feel, I have recently rediscovered how effective the sharpness equalization tool really is and I find myself using it more often. Here is an example that shows how detail such as red/blue can appear soft compared to B/W detail in the same focal plane and how sharpness equalization can help resolve problems of sharpness and depth:

Original After sharpness EQ

Notice how the color detail (particularly the red) in the image on the left appears softer than the B/W detail in the upper left quadrant. This is due to the sensor having less information to work with when capturing saturated colors. The red detail in the image on the left almost looks like it is in front of (or behind) the B/W detail due to the red detail being a bit out of focus. In reality, this is a test target on a flat sheet of paper so all of the lines in each quadrant should have the same sharpness. Take a look at how sharpness equalization has corrected this on the right image. The color (red, green, and blue) detail is now just as sharp as the B/W detail in the upper left quadrant. The sharpness equalization has now effectively restored sharpness in the photo and along with it the proper depth of field. To see examples of how this works with real photos, see my earlier article from Digital Outback Photo or download a trial of my own Qimage batch printing/processing software and look in the help under unsharp mask to see how you can try this process on your own photos!

Summary

If you're like me and you want to get the most detail out of your photos but you always find something missing when capturing bright colors, take a look at the information in this article. You may be noticing a discrepancy in sharpness/detail produced by your camera due to the way your camera captures color. Using sharpness equalization can help you gain more "3D effect" or feel from your photos and increase the overall presence of the scene.

-- Mike Chaney