Someone explain the science to me on this - cuz I don't get the blue shift - as you see in this test: the GH2 does not reproduce colors accurately in blue light. So are we skewing blue, or are we skewing yellow?

I will shoot color charts using different gels to see how it affects the sensor - then shoot the same charts with custom WB through filter and do the same so we can see what's happening. I want to know exactly what's happening instead of conjecture. But it may actually be a way to get better images.

I know this just from one day; putting the minus green gel (that's what CC30 is in Lee filter terms) between the lens and the sensor softens the image up nicely without blurring. So just for that, I like it. But I think it's just because my gel insert isn't perfectly clear but slightly scuffed. And there's less contrast in the image.

I saw something I liked in some footage shot with the CC30 magenta behind the lens in front of the sensor, enough to want to play with it.

Let the CC30 filter test begin. I have one mounted in front of the sensor and am playing with it. I like what I see so far. Will post when I've shot enough footage in enough different situations to be relevant.

Gonna be difficult though I have two preset manual white balances for daylight and tungsten with the filter mounted, I guess I could set the other 2 for it without the filter and also shoot with the filter without WBing the image and correcting in post. Before I waste a bunch of time, any suggestions on what you want to see?

For the past few days I have been working on lenses previously uncommon to me, and I did not have a suitable 67 mm circular magenta filter. So it was back to filming without boost in saturation and flatter contrast. No surprise to me the colors looked lame and shadows noisier and murkier.

So I put a Kodak CC20M filter inside a FD-to-M43 adapter and this seems to work nicely. I also used a CC20C (cyan) for white balancing. Now the colors "pop" (image attached). When the saturation is boosted to this kind of levels, the 4:2:0 nature of the video becomes apparent, as seen in the image. Especially transitions from blue to red are coarse. This could be easily softened, but I left it there in the image to show what I mean. It would now make sense to record in 4:2:2, if possible, and that would be one step closer to really filmic colors.

I have turned this color balancing method on location into a set of simple tools that can be put in the pocket and use easily to reliably white balance for certain boost in saturation.

I just moments ago snapped the attached image outside when sun was just rising.

The left sample is straight from camera taken as a jpeg still image (lowest quality setting). White balancing was done using yellow filter, then I removed the yellow filter and installed CC30M magenta filter. It looks rediculously balanced, but on the right is the same image with colors adjusted in Photoshop. Not so rediculous any longer. In this case I had quite a bit of freedom to adjust for the yellow - blue slider. If I needed a good strong unbanded sky I could have tried to balance for the sky using a cyan filter, for example.

It is an extreme very quick example and balancing, but came out with rather filmic colors and contrast. Vertical stripes are caused by GH1 fixed pattern noise. I believe that this technique when mastered holds the key for the best and most filmic graded video from GH cameras. Lens is the Rokkor 45 mm @ f2.

easy Aki. You are doing a grate job. Just keep on posting the images and videos side to side, so people can compare. Just make always available the original MTS so anyone can make its judgment.

@aki_hartikainen -

http://en.wikipedia.org/wiki/Bayer_filter

"I did not write that."

That's reassuring to hear. I'd suggest you read it more closely, as there's nothing in that link pertaining to 4:2:2 color sampling.

@aki_hartikainen - "Bayer sensor is inherently 4:2:2 whereas film is 4:4:4. "

No, you're confusing YUV sampling with Bayer sensor resolution and coming up with nonsense. Please stop spreading misinformation.

I could see, but the results are getting rather good and filmic as understanding of dealing with limited color Bayer sensor and limited latitude and color supressed codec improves. Bayer sensor is inherently 4:2:2 whereas film is 4:4:4. By adding magenta (Blue + Red) the quantity of color atleast is fixed to the same relative amount as for film.

I am still on the GH1 which is a good camera but not fantastic. Fixed pattern noise could ruin otherwise good and moody low light scenes.

It appears two additional CC filters might be handy. Those would be yellow and cyan. For example, if you needed to pick bluish or yellowish hues in secondary color correction and had to boost those, it would help if the blue channel had been captured on location with the widest latitude. You could try this;

White balance the camera for white having a yellow filter in front of the lens. Remove the yellow filter and attach magenta filter. Expose the image so that the blue channel has full range. Examples below. Now the blue - yellow slider will work very well in editing program and one could pick both bluish and yellowish hues with ease. Yellow high lights will bloom just like they should. Yellow hues are captured by the blue sensor, they are just phase shifted. Likewise, if you had to pick reddish or cyanish hues it would be good idea to make sure red channel is exposed for full range. So using a cyan filter (opposite of red) when white balancing would accomplish this.

Good brown hues would depend on whether they are "yellowish brown" or "reddish brown". This way we could optimize the limited palette of the camera for grading. So the grading actually starts in the camera for best results.

Also, for film negative the primary colors are cyan, magenta and yellow (the opposite of primary colors for video). So one could see we have a filmic approach to the colors with these CC filters. As always, I am trying to create a bit better article about all this.

I learned that there is a program freely available called WinFF. It is a graphical user interface for ffmpeg conversion program that you can use to batch convert MTS files to AVI files or image sequences. So no need to start coding new software. ffmpeg does an excellent job converting MTS files.

There is plenty of new information for Project Blue Light worth a mention. I have made several videos using ffmpeg to convert first to HuffYUV AVI files. I edited the video as usual in HuffYUV and exported the edited video into image sequence to be graded in Silkypix. This way it is possible to save the grading as the last step and really massage the colors to one's liking.

One such video is now on-line, if it is okay to link to it, here it is:

This is a small video filmed indoors in diverse lighting conditions, and it came out really fine. Even the YouTube version is largely free of banding and posterization and does not immediately look like it was filmed with a eur 300 Lumix GH1 and eur 100 Tokina 24 mm lens. Worth to have a look.

Other mentions are about the lenses to use with magenta filter. I have noticed that some lenses work better with magenta filter than others and the images are more easily balanced to one's liking. An excellent lens hood is mandatory item to avoid purple haze that lowers contrast and is difficult to remove in post. I found an old big lens hood padded with non-reflecting material and this clearly improved the image (however not yet used on the video linked above).

I plan to post an update about all this on my website hopefully real soon.

Did you see the banding during playback ? What player did you use ?

Aki.. Would you consider trying the same workflow using a magenta shift in-camera ? No color filter, just use the WB settings.