Quite a bit has happened since starting this topic. If you want to cut to the chase and want the quickest way to eradicate focus pixels simply download MLV App and load in the latest focus pixel map files. This video shows how to install the map files into the app:
The rest of this is post for historical (hysterical?) purposes:
@dfort,
One possibility is mapping the focus pixels for each camera and then interpolating out those pixels (rather than just chroma-smoothing the entire image). This might fix the dual ISO / focus pixel issue, IDK. It should be faster than chroma-smoothing, and not have any artifacts. It's a lot of work though, you need to create a focus pixel map for each camera and video mode.
So here we go--Affected cameras: EOSM, 100D,
600D, 650D, 700D, (and more?)
[EDIT: Pretty sure now that the 600D doesn't show focus pixels in raw video.]
Focus pixels, phase detectors, call them what you will when these dots show up on your raw video they are annoying. There have been attempts at clearing them up like the
PinkDotRemoval tool that has apparently been abandoned. Chroma smoothing helps but it affects the overall image and it doesn't work for dual iso footage.
I tried to shoot something that shows as many of these pixels as possible by shooting an out of focus white paper with a couple different colors of lights and processing the image with a heavy hand to show the pixels as clearly and dramatically as possible.
Got your attention? This is a frame from an mlv file shot on an EOSM in crop movie mode. It doesn't look quite as bad when shooting regular subjects but it is a problem.
It is possible to get an even clearer view of these pixels by running a frame through dcraw using the -d option which, according to the documentation:
-d
Show the raw data as a grayscale image with no interpolation.
Here is a closeup of that same frame at the top of this post.
Once we know the coordinates of these pixels they can be zapped.
Method 1 - In CameraThe ideal way to deal with them would be in the camera. In fact there is a way that "bad pixels" can be mapped out and apparently it is common practice to map out these faulty pixels when the camera is manufactured or repaired--saves on reject sensors. Assuming that we can't access the code that maps out pixels in the Canon firmware, the developers over at the chdk project have figured out more than one way to do it:
http://chdk.wikia.com/wiki/Badpixel_removalI'm not skilled enough to take what they have done and apply it to Magic Lantern but maybe someone who has the knowledge can drop a few hints?
Method 2 - In PostYeah, we can fix it in post. In fact there are several programs that can deal with pixel problems.
Pixel Fixer is a dedicated tool,
RawTherapee has an option to deal with hot and dead pixels and an interesting Python wrapper for the LibRaw library called
rawpy that creates a ".badpixels" file dcraw can use to zap out pixels. This might be a way to create the database of focus pixel maps dmilligan suggested at the top of this post.
I tried using dcraw to get rid of one of the focus pixels on a dual iso frame and found that one pass wouldn't do it. Maybe it is because the process is breaking down by running the file through cr2hdr first? My understanding is that the debayering process is averaging neighboring pixels and in dual iso it seems to "smear" the pixel vertically. In addition, some of these focus pixels have a secondary pixel diagonally across that also needs to be mapped.
Here is what I ended up doing to get rid of just one focus pixel--and the diagonal secondary pixel.
1247 121 0
1248 120 0
1248 121 0
1248 125 0
1248 126 0
1248 122 0
1248 123 0
1248 124 0
1248 126 0
1248 122 0
1248 120 0
1248 124 0
Note that I hit the diagonal pixel just once but had to work top and bottom to middle and hit the same pixels several times in different sequences. Think of it as stippling a brush to spot a print--in photography school I spent many hours spotting prints. (By the way,
this a map of all the focus pixels for the 650D that was used on the PinkDotRemover tool.)
Here is the result. I circled the area where the hot pixel was located and yes, it looked just like those other ones before running the dng through dcraw.
So if this is the solution then what's the problem?
The problem is that I'd like to get some feedback that this would actually be practical. It looks like there would have to be a different focus pixel map file for every camera and every crop factor (raw image size) and all of this would have to be assembled in a way that users could handle. Danne is the master of assembling scripts that can automate complex processes so it is possible to piece together something with dcraw--but then again is that the best solution? Maybe it would be better to adapt dcraw's "badpixels" code to mlv_dump or cr2hdr or MLVFS?
Just asking for a little guidance here.
