[600D RAW] Comparing Tragic Lantern [NewMem] vs. [New System]

[vickersdc]

Really quick comparison test of different versions of raw recording on the 600D.

I started using the Tragic Lantern [New System] back in June, and then updated to [NewMem] late November (2013). I've been plagued by multiple dots in the image when using [NewMem], that I didn't recall seeing in the older [New System].

So I ran a quick check... there aren't any dots in [New System] as you can see in this video. Take a look at the upper centre of the first clip, in the sky area, to see the dots. Same camera, same lens, same workflow (RAWMagic -> Davinci Resolve) for both raw recordings... interestingly applying the exact same 'look' to both recordings resulted in a different response by the clip.

To be fair, on another check I did, [New System] did suffer from some banding in the image, but I need to do further checks on that.

I'm about to go and do some lowlight checks with [New System] and see how that looks - if it's OK, I may just go back to that raw recording.

[a1ex]

This was discussed back then, in that prehistorik era.

Bottom line is that on some cameras you can have black dots (easy to detect and remove, e.g. in ACR) or autofocus dots on high-contrast subjects (less noticeable, but impossible to locate and fix). Implementing a fix for black dots in raw2dng is very easy and takes 5-10 lines of code (hint, hint).

[vickersdc]

Hi a1ex,

I know ML / TL moves fast, but is June prehistoric? On second thoughts, it probably is 

I don't have ACR, only FCP X and Davinci Resolve 10 Lite. I did do some searching of other posts about black dots, and  came across a pretty recent one that said this hadn't been fixed. I'm now wondering if they meant that it hadn't been sorted within the raw module.

5-10 lines of codes, eh? It's been a long, long while since I did any coding (20 years ago)! I've found plenty of posts of green dots, purple dots, etc. but they all seem to require Adobe-something. I'll search again for dots and raw2dng and see what I can find... any hints on the coding required?

DV.

[a1ex]

Yeah, just copy it from dcraw or ufraw.

[vickersdc]

Am I on the right lines... do I need to create a .badpixel file listing the x,y co-ordinates of the black pixels? Then use RawTherapee, or whatever, to overwrite those pixels based on the values of the pixels around it?

That's as far as I've got so far 

[a1ex]

No. Look up the previous work done by Marekk on this. I would undo the badpixel opcodes approach (only recognized by ACR) and replace them with a interpolation routine (you can grab a median filter from cr2hdr cold pixel fix, or the dcraw/ufraw approach which seemed a bit simpler last time I've looked at it).

(I will also need removing the bad pixel approach because I want to upgrade to DNG 1.4 to be able to write floating-point DNGs)

[vickersdc]

I've searched for marekk's profile and found his 8 pages of posts - to be honest I'm thinking dealing with the black dots issues is way beyond my skillset! The bottom line is that I really don't know / understand what to do.

a1ex: I'm very grateful for your hints, but I'm stuck. I'm also really aware that asking for assistance when the answer has already been given on the forum is not a way to endear yourself to the developers / helpers... but I've searched and searched! It seems that others who are also running Resove are having the same issues. I've tried different versions of raw2dng, and RAWMagic, and on Windows and Mac systems... the same output on all of them.

Raw recording is just so good, but with the black dot issue, it means that I can't really use it - and that's such a hard decision as the workflow is good, the grading is great, it's sharp and I enjoy working with it. But if I can't get a solution (and I have spent hours looking) then I'll have to either go back to the [New System] build or revisit raw in the future.

Can anyone help? I'm using Mac OS X 10.9 (Maverick), Davinci Resolve 10 Lite and FCP X 10.1 (if that's any use).

Thanks,
David.

[a1ex]

Come on... for each pixel, if it's black (value=0), let it be the average (or median or whatever) of its neighbours. Code it into raw2dng. That's all you need to do.

[vickersdc]

I took a closer look at these 'black' dots, and they're not black at all.

I swapped over to MLV recording and then used mlv_dump in various ways to see what happened with converting to .raw, converting directly to .dng (using each option for chroma smoothing) and found that the best results were with direct to .raw, or .dng (no chroma smoothing).

Anyway, when I looked at the dots this is what I saw...



There are many, well over 20, of these spread throughout the image, and whilst the central cluster of four pixels can be in any arrangement (i.e. the darkest one isn't always bottom right), that cluster is always surrounded by brighter looking pixels.

So, just looking for a black pixel isn't going to cut it... any ideas?