XDCAM cameras have Sony’s Skin Tone Detail Correction system included in the picture profiles. By turning this on you can point the camera at a face (or any other coloured object) and select the hue you want to treat. By using the phase and saturation controls you can adjust the exact hue and hue range that will be treated. Then you can turn the detail level up and down for the selected range.
It works but is a little fiddly to set. I don’t normally use it, instead preferring to shoot with slightly reduce detail level settings overall and then adding a diffusion filter in post production using Magic Bullet or similar. Another option would be to use a diffusion filter or similar on the camera, I like the Tiffen Gold Diffusion/FX for faces. If your budget won’t stretch to that then don’t forget that you can always stretch a very fine mess net over the lens such as a stocking for a pleasing diffusion effect. Again tricky to get just right, if the mesh is too big you’ll see it, too small and you completely blur the image.
Why do my pictures go soft when I pan? Camera Detail Correction in depth.
This article is my Christmas present for my readers. When your trying to set up a camera or brew up a picture profile it really helps if you understand the ramifications of each of the settings. I hope this helps explain how detail correction works and how it effects your image.
I am often asked to explain why someones images are going soft when they pan the camera or when there is a lot of movement in the scene. Well this can be down to many things including poor compression or too low a bit rate for the recording, but the two main issues are shutter speed (which is tied in to your frame rate) and detail correction. I’ll cover frame rates and shutter speeds in the near future, but today I’m going to look at Detail Correction.
First of all what is detail correction for? Well originally it was used to compensate for the low resolution of domestic cathode ray tube TV’s and the limited speed at which a CRT TV could go from light to dark. Modern LCD, Plasma and OLED displays handle this much better, but still detail correction remains important to this day to as a way of adding the appearance of additional sharpness to a video image. You’ll often see extreme examples of it on SD TV shows as a dark halo around objects.
Perfect Greyscale
The image above is of an imaginary perfect greyscale chart. Looking at it you can see on your screen that each grey bar is quite distinct from the next and the edge between the two is sharp and clear. You computer screen should be quite capable of showing an instant switch from one grey level to the next.
Grey Scale with Waveform
Now if we add the waveform that the “perfect” greyscale would give we can see that the transition from each bar to the next is represented by a nice crisp instant step down, the transition from one bar to the next happening over a single pixel.
Grey Scale as seen by camera
The image above represents what a typical video camera might reproduce if it shot the greyscale chart without any form of detail correction or sharpening. Due to the need to avoid aliasing, lens performance and other factors it is impossible to get perfect optical performance so there is some inevitable blurring of the edges between the grey bars. Note that these images are for illustration only, so I have exaggerated the effect. I would expect a good HD camera to still produce a reasonably sharp image.
Camera Waveform
Looking at the cameras waveform you can see that the nice square edges we saw in on the perfect greyscale waveform have gone and instead the transition from bar to bar is more rounded. Now there are two things that camera manufactures commonly do to correct or compensate for this. One is called aperture correction which is a high frequency signal boost (I’ll explain that another time) but the one were going to look at in this case is called detail correction often simply referred to as “Detail”.
Detail Correction sampling
So what happens in the camera? Well the camera constantly compares the video luminance (brightness) levels of the image over a set time period. This time period is incredibly short and in the example given here is the time it takes for the cameras line scan to go left to right from point A to point B. If the difference in the brightness or luminance of the two samples is greater than the threshold set for the application of detail correction (known as crispening on Sony cameras) then the detail circuit kicks in and adds a light or dark enhancement to the brightness change.
Camera image with Detail Correction added
With an HD video camera the light or dark edges added by the detail correction circuit are typically only a few pixels wide. On an SD camera they are often much wider. On a Sony camera the detail frequency setting will make the edges thicker (negative value) or thinner (positive value). The Black and White limit settings will limit how bright or how dark the added correction will be and the detail level control determines just how much correction is added to the image overall.
One important thing to consider is that as the amount of detail correction that is applied to the image is dependant on differences in the image luminance measured over time, so you have to consider what happens when the scene is moving or the camera pans. Two things happen when you pan the camera, one is that the image will blur a little due to things moving through the frame while the shutter is open and from line to line objects will be in a slightly different position.
Blur due to camera pan softens the image.
So looking at the waveform we can see that the waveform slope from one grey bar to the next becomes shallower due to the blur induced through the motion of the camera. If we now sample the this slightly blurred image using the same timescale as before we can see that the difference in amplitude (brightness) between the new blue samples at A and B is significantly smaller than the difference between the original red sample points.
Smaller Luma difference due to pan blur
What this means in practice is that if the difference between the A and B sample drops below the threshold set for the application of detail correction then it is not applied. So what happens is that as you pan (or there is motion in the scene) the slight image softening due to motion blur will decrease the amount of detail correction being applied to the image so the picture appears to noticeably soften, especially if you are using a high detail correction level.
Detail correction is applied to both horizontal image differences as outlined above and also to vertical differences. As the vertical sampling is taken over 2 or 3 image lines there is much longer time gap between the samples. So when you pan, an object that was in one position on one line may have moved significantly enough by the time the frame scan has progressed 2 more lines that it is in a different position so the detail sampling will be wrong and detail may not be applied at all.
If you are finding that you are seeing an annoying amount of image softening when you pan or move your camera then you may want to consider backing off your detail settings as this will reduce the difference between the detail “on” look and detail “off” look during the pan or movement. If this softens your images too much for your liking then you can compensate by using Aperture Correction (if your camera has this) to boost the sharpness of your image. I’ll explain sharpness in more depth in a later article.
After completing the multi camera shootout at Visual Impact, one thing was bothering me about the pictures from the PMW-350 and that was the way the specular highlights in the tin foil were artificially enhanced. During the test the camera was set to factory defaults, which IMHO are too sharp, but the foil in particular looked nasty. Since then I have been further refining my paint settings for the 350 and looking at detail and aperture. Today I was replicating the tin foil test and looking at the aperture settings (not the knee aperture) and I noticed that turning aperture on and off had a very pronounced effect on highlights but a much smaller effect elsewhere in the image. Normally I would expect the aperture setting to act as a high frequency boost making subtle textures more or less enhanced, which it does, but the amount of enhancement appears to vary with the brightness of the image with specular highlights getting a really big hit of correction. If you look at the images to the left at the top you have aperture correction on at +99. There are big ugly black lines around the highlights on the foil and the texture of the carpet has been enhanced. To some degree this is the expected behaviour although I am surprised by how thick the edges around the highlights are, this looks more like detail correction (it could be “ringing”). The middle images are aperture off, not zero but actually off and you can see that the edges on the foil have gone and the carpet is no longer enhanced. The bottom picture though with aperture on at -99 though is very interesting as the carpet appears slightly softer than OFF, which is not unexpected while the foils is sharper than OFF and this is not expected. I don’t like this behaviour I’m afraid to say as a typical way to get a filmic look from a video camera is to turn the detail correction off to give a natural picture and then use Aperture correction to boost high frequencies to retain a sharp image. On the PMW-350 you can’t do this as this as a high Aperture setting will give you those nasty edges on highlights. So what can you do? Well the 350?s native, un-enhanced resolution is very high anyway so it doesn’t need a lot of correction or boosting. The default Detail and Aperture settings will give some really nasty highlight edges so you need to back things off. If your going for a filmic look I would turn OFF aperture correction altogether, for video work with pictures that have some subtle enhancement I would use Aperture at around -20, certainly never higher than -15 unless you like black lines around specular highlights.
My current prefered detail, aimed at giving a very slight, not obvious enhancement are are as follows:
Detail Level -12, H-V Ratio +15, Crispening 0, Frequency +30, White Limit +30, Black Limit +40 (all other detail settings at default)
Aperture OFF for filmic look, Aperture -20 for video look.
I have also made some changes to the Matrix settings. I have been finding the pictures from Sony cameras to be a little on the Green/Yellow side so I have tweaked things a little to remove the yellow cast and put in a bit of red, this is a subtle change but really helps with skin tones, stopping on screen talent from looking ill! These settings work in the PMW-350, EX1/3 and PDW-700.
On an EX1/EX3 this works best with the Standard Matrix, On a PMW-350 or PDW-700 you can use it on it’s own or mix it with one of the preset matrices as a modifier. User Matrix On, R-G 0, R-B +5, G-R -6, G-B +8, B-R -15, B-G -9
I got given the link to a video on the RedGiant site (thanks Paul) that gives a really interesting insight into how the grading in blockbuster movies is done. The nice thing about the video is that it explains why certain things and certain colours are used. It’s a really insightful video.
