I kind of feel like we have been here once before. That’s probably because we have and I wrote about it first time around.
A typical video camera has a special filter in it called an optical low pass filter (OLPF). This filter deliberately reduces the contrast of fine details in the image that comes from the cameras lens and hits the sensor to prevent aliasing, jagged edges and moiré rainbow patterns. It’s a very important part of the cameras design. An HD camera will have a filter designed with a significant contrast reduction on parts of the image that approach the limits of HD resolution. So very fine HD details will be low contrast and slightly soft.
When you shoot with a 4K camera, the camera will have an OLPF that operates at 4K. So the camera captures lots of very fine, very high contrast HD information that would be filtered out by an HD OLPF. There are pro’s and con’s to this. It does mean that if you down convert from 4K or UHD to HD you will have an incredibly sharp image with lots of very fine high contrast detail. But that fine detail might cause aliasing or moiré if you are not careful.
The biggest issue will be with consumer or lower cost 4K cameras that add some image sharpening so that when viewed on a 4K screen the 4K footage really “pops”. When these sharpened and very crisp images are scaled down to HD the image can appear to flicker or “buzz”. This will be especially noticeable if the sharpening on the HD TV is set too high.
So what can you do? The most important thing is to include some form of anti-aliasing to the image when you down scale from 4K to HD. You do need to use a scaling process that will perform good quality pixel blending, image re-sampling or another form of anti-aliasing. A straight re-size will result in aliasing which can appear as either flicker, moire or a combination of both. Another alternative is to apply a 2 or 3 pixel blur to the 4K footage BEFORE re-sizing the image to HD. This seems a drastic measure but is very effective and has little impact on the sharpness of the final HD image. Also make sure that the sharpening on your TV is set reasonably low.
I previously wrote about this very same subject when HD cameras were being introduced and many people were using them for SD productions. The same issues occurred then. Here are the original articles:
The C300 is not Moiré free as can be seen from this blown up section of a frame grab. Once again it’s fine brickwork thats causing the problem. Now before everyone runs off in a panic, lets put this into perspective. The F3′ aliases, the Alexa aliases as do most single chip cameras. This is certainly no worse than an F3 and is right at the resolution limits of the camera, so your not going to see it very often. It takes a very fine, high contrast pattern, in sharp focus before you’ll see this kind of thing.
This a recurring question that I get asked about time and time again. The main problem being that the SD pictures, shot with an HD camera look soft. So why is this and what can be done about it?
Well there are several issues to look at. First there is camera optimisation. Sadly what works for HD doesn’t always work well for SD. Secondly there is the downconversion process. If your shooting HD and simply outputting SD using the cameras built in downconverter than you really don’t have many options but if your using a software downconverter you may be able to improve the results your getting.
Starting with the camera, what can you do? Well first off let me say that a camera optimised for HD will always be a compromise when it comes to SD. As the native resolution of HD cameras increases then the problem of getting good looking SD actually gets worse. The problem is that a good high resolution camera will normally only have a very small amount of artificial sharpening via the detail or aperture circuits, because in HD it will look nice and sharp anyway. SD cameras and the SD TV system with it’s inherently low resolution and soft pictures has always relied very heavily on detail enhancement to try and make the pictures appear sharper than they really are. When you take the minimal additional sharpening of an HD camera and downconvert it to SD it all but disappears, the end result is a soft looking picture. There is no easy fix for this, you can either add additional extra thick detail correction edges to the HD pictures, which risks spoiling the HD image or you can add additional detail correction in post production. On a Sony camera the thickness of the detail correction edges is controlled using the “frequency” setting. Setting this to a negative number will thicken up the detail edges, very often you need to go all the way to -99 to get an appreciable difference. As an alternative you can add extra sharpening or detail correction in post, after the downconversion process. This is the way I would go whenever possible as I don’t want to compromise my HD pictures for the sake of the SD images.
The second issue is the quality of the downconversion. A simple rescale from HD to SD rarely works well as it can create a lot of aliasing. Aliasing is the result of taking too much detail and trying to record or represent it with too few pixels. See this article for more on aliasing. Imagine a diagonal line running through your image.
If you sample it at a high resolution, with your HD camera then the line looks reasonably good as you can see in the diagram to the left.
If you then take that HD captured edge and simply scale it down to SD, you quarter the number of samples and the end result is a jagged, stepped line. Not pretty. In addition, if the line moves through the image it will flicker and “buzz”. This is far from ideal.
A better approach is to blur the HD image before down converting using a 4 pixel (or similar) blur, or to use a downconversion programme that will include smoothing during the conversion. The final image shows the kind of improvement that can be gained by softening the image before down conversion. The blur around the edges of the line soften it and make it appear less jagged. This will result in a much more pleasing SD image. Next you then add in some detail correction to restore the apparent sharpness of the image and viola! A decent looking SD image from an HD source. In compressor to get a good downconversion you need to activate the advanced scale tools and use the “better” or “best” scaling options.
Here is a roughly done (sorry) comparison of the aliasing from an EX1R at the bottom and my F3 at the top. The F3 had a Nikon 18-135mm zoom, both cameras were set to default settings, 25P. The F3 clearly shows a lot more chroma aliasing appearing as coloured moire rings in both the horizontal, vertical and diagonal axis. The EX1R is not alias free. The chroma aliasing from the F3 is not entirely unexpected as it has a bayer sensor and there is always a trade off between luma resolution and chroma resolution and the point where you set the optical low pass filter. Frankly I find this performance a little disappointing. More real world test are needed to see how much of a problem this is (or is not). To put it in to some perspective the F35 aliases pretty badly too, but that camera is well known for producing beautiful images. I hope I’m being over critical of this particular aspect of the F3’s performance, because in every other respect I think the camera is fantastic.
UPDATE: I’ve taken a look at the MTF curves for the F3 and they are quite revealing showing that an OLPF is in use which is giving an MTF50 of around 800 LW/PH V and 950 LW/PH H. This is not quite as high as an EX1 and are quite reasonable figures for a 1920×1080 camcorder. This suggests that the aliasing is largely limited to the chroma sampling of the sensor. As this is a bayer (or similar) type sensor the chroma is sampled at a reduced rate compare to luma, which is why coloured moire is not entirely unusual.
Tests performed with a Tokina ATX-Pro 28-70mm lens at 25P
I’ve been testing and playing with my F3 and first off let me say this… I love this camera, it produces amazing images and I can play with lots of lenses!
But, it’s not all roses. The F3 suffers from aliasing. A zone plate shows extensive aliasing. Until today this had not caused any concerns as I had seen little evidence of it in actual footage, but when shooting some brick houses this afternoon I came across some coloured moire patterns appearing as faint coloured stripes across the brickwork in the footage. It’s not anywhere near as bad as a DSLR, my wife looked at the footage and didn’t notice it until I pointed it out to her, but it’s certainly there. This is disappointing on a camera at this price level, my EX1 doesn’t do this. Now the zone plate shows this to be an issue with the cut off of the optical low pass filter, so I doubt that there is much that can be done in the firmware, but then Sony have done some clever stuff in the past with firmware updates. When working on my picture profile settings I did find that increasing detail above -15 would increase the visibility of the aliases on the zone plate, however when I tested various detail settings with the brickwork there was little difference. I think Nigel Cooper has also seen this, but I’ve not seen it mentioned anywhere else. Has anyone else observed this?
As you may have seen from my earlier post I became the owner of the new Canon T2i (or 550D as it’s known in the UK) at the weekend. Clearly before using any camera in anger it’s important to see what it can and can’t do. I will say that I am not a Canon DSLR expert. I have been following the fuss and much admire some of the work done with these cameras by Phil Bloom, but frankly after playing with the Canon over the weekend I have to say I’m disappointed. Yes you can achieve shallow depth of field very easily and you do get a filmic look to the pictures, but look at the footage on a big monitor and it just looks soft. At first I wondered if this was the lens I was using, so I tried a couple of others including a nice Tamron 28mm prime. I tried different apertures, shutter speeds etc, but every clip I’ve taken looks soft. In isolation, on scenes with low detail this isn’t immediately apparent, but anything with lots of fine detail looks soft. Some of this is aliasing, look at the roof of the house in the T2i image, it appears to have diagonal roof tiles, this is a pretty typical aliasing artifact. I shot some closer shots of the buildings and the brickwork aliased like crazy.
Looking at the flowers picture you can see that the EX1 has picked up more of the subtle texture, or at least it has recorded more of the texture. I’m sure some of the Canon’s softness is due to compression artifacts. The other thing that I found is that it is tending to crush blacks a bit. I have played around with the picture styles and you can reduce this a bit, but there is very little detail in deep blacks, which would IMHO make grading tricky. The one good thing I did find was that it is very noise free at 200 and 400 asa, it’s also useable up to 800 asa or at a push 1600asa, so it would make a good camera for very low key scenes, provided you use a good fast lens. Looking at the Canon pictures there was something pleasing about the deep, almost crushed blacks. I think this helps contribute to the Canon DSLR “look” so I quickly threw together a new picture profile for the EX1/3 and PMW-350, but I’m afraid that the details of that will be the subject of another post, as I have work that I must do first! The EX images in the frame grabs were shot with this picture profile. As we all know the ergonomics of the video DSLR’s is pretty poor for video. It’s tricky to hold and you have to use an add on Loupe to make the LCD useable as a viewfinder. You can’t zoom mid shot and without peaking or zebras adjusting exposure and focus accurately is difficult. I was hoping to be able to use the 550D as a B camera for those situations where I need a small, discreet camera, but having seen the pictures, so far, for me it will be reserved for holidays and shooting where you not supposed to video and for shoots where supper shallow DoF is essential. I have to say I’m really disappointed, I wanted this camera to be so much better, I knew it would suffer from aliasing, but I wasn’t expecting the soft pictures, I guess some will say that the softness adds to the filmic look, but I’d much rather do that with some nice pro-mists or filtration in post production rather than starting out with soft pictures. Perhaps I’ve done something wrong? If I have please add a comment!
UPDATE: I was so convinced that I must be doing something wrong that I shot some more clips, this time with less harsh lighting. No, change however, the T2i is still soft and the new clips show just how big a problem aliasing is. You have to consider that the coloured moire patterns are recorded like that, no amount of grading will get rid of it. A small amount of diffusion on the camera should help, but then your going to have to work out how much to soften and diffuse each shot to make sure your not making the pictures even softer than they already are.
The frame grabs are all 1:1 pixel for pixel, no trickery has been used! You can download some further examples by clicking here. Even if you were shooting stuff for the web this level of aliasing could cause big problems as it’s really obvious. For this shot I had the Canons sharpness setting turned all the way down. I have also turned down the contrast setting as this gives better dynamic range with less crushed blacks. My workflow is to import the H264 files from the camera and then convert them to ProResHQ. This helps a little with sharpness over working with the native H264, but for me this last test was the nail in the coffin for DSLR’s as footage like this would simply be unusable. If you watch the YouTube clip please make sure you watch it full screen or at least at the 480P setting. The small embedded size doesn’t show the aliasing as much as the bigger versions.
OK, so it’s defiantly not just me doing something wrong. When in focus the T2i/550D aliases (as do all the current Canon DSLR’s). This is a grab from Philip Blooms latest Canon short. For once this is a daylight piece and as I expected it exhibits a lot of aliasing. The grab is actually taken from the thumbnail on his exposure room page. I’m really pleased to see this as it shows that aliasing is a problem for the experts too. You start to appreciate why so many of the Canon shorts are shot at night, with millimeter deep DoF… it’s to stay clear of having stuff in focus that will alias. there are filters from Caprock that are supposed to help, but you need a different filter for each focal length and aperture that you use, they also soften the picture somewhat.
If you want my opinion, then it has to be that the Canon’s are close, but still a mile away. The aliasing issue is a biggie. Sort it out and the skew, jello and overheating can be worked around, but if you have to worry about simply having a piece of wood in focus and whether it’s going to exhibit rainbows of colour or whether cobble stones will twitter and change colour (At 00.35 and this is from Canon) then it will limit what you can do. There is quite a lot of aliasing in Phil’s new daytime clip, basically anytime anything is steady, has texture and is in focus, it aliases. I’ve been shot down in flames on other forums for saying that this is a problem, but if even the experts can’t deal with it then what hope does everyone else have? I would love to have the option of shooting with the shallow DoF that the Canon’s offer, but not at the expense of having to avoid any kind of texture. Perhaps Red and Scarlet will be better, perhaps Canon will sort it out, or perhaps not, as the cameras are clearly selling like hot cakes, even with the issues. If they do fix it then the camera will almost certainly be for video only.
Ever since the release of the XDCAM EX cameras users have been having problems getting good looking SD pictures out of downconverted HD.?Why is this and what can be done about it? This is an issue that effects all high resolution HD cameras and is not unique to the EX’s. There are two key issues. The first is the way basic software converters handle fields in interlace material and the second is the amount of information in an HD image that must in effect be discarded to get a SD image.?At first glance you would think that starting off with lots of picture detail would be a good thing, but in this case it’s not. Let’s see if I can explain.?Imagine that you have something in you HD picture that over 4 pixels goes from light to dark, in Hd you get a gradual transition from light to dark and all looks good. Now what happens when you take those 4 pixels and convert them to SD. The 4 pixels become just 2 and instead of a stepped change from light to dark the picture now goes instantly from a light pixel to a dark pixel. If these pixels were the edge of a moving object, as it moved the pixels would be switching instantly from on to off and unless the object moved at exactly one pixel per frame you will get a flickering effect. Clearly our nice gradual transition from light to dark has been lost and if there is any motion we may now be seeing flickering edges. Niether of these look good.
Take a look at these images:
As you can see the down converted SD is very blocky and there is some strange patterning (aliasing) going on amongst the bricks of the houses in the background. This does not look good and if there was motion the brickwork would shimmer and flicker.
So what can be done?
Well the best way to improve the SD down conversion is to soften the HD image before it is down converted to prevent this single pixel light to dark switch from happening. You need to end up with an SD image where you go from full light to full dark over at least 3 pixels to prevent flicker (Twitter).
How much you will need to soften you HD by will depend on how sharp it is to start with. Simply turning down the cameras detail settings can be a big help, but even then the best results are often obtained by applying some kind of blur filter in post production. In FCP I find the flicker filter works quite well. As you can see from the frame grab below the difference in the quality of the downconvert is quite striking.
I have also found that another problem is that the detail settings on an HD camera are not optimised for SD. The detail correction edges created in HD are very thin and when these are down converted to SD they all but disappear and can cause further aliasing. The solution is to make the detail correction edges thicker (on an EX turn detail frequency down to -60 to -99) but this then looks ugly in HD. The bottom line is that a camera optimised for HD works best in HD and SD will be a compromise.
Camera setup, reviews, tutorials and information for pro camcorder users from Alister Chapman.