![fastrawviewer 1.3.3 fastrawviewer 1.3.3](https://macupdater.net/cgi-bin/icons/_default_app_icon.png)
Let's check some sample data obtained from ColorChecker bottom row under different light (Git2P, RAW, no OB subtracted, 8 bit): Buuut, at least one tool blindly applied making ugly default image, so not sure if this is safe.Īnother interesting thing about Black Level (OB) is its connection with White Balance.
#FASTRAWVIEWER 1.3.3 SOFTWARE#
In most software this forces autowb and the result is mostly superior to small functions. 0xc628 tag not present (exiftool can erase it). It is worth looking what open source tools (ufraw./darktable/.) implemented, maybe there is a neat small function.Īh yes: another idea about WB. That is also a reason why I asked audience for some colorful test images. Looks promising and bit more accurate than gray world. I have cooked a version now of a "specular reflection" idea (the brightest pixels indicate wb). Gray world (which you use?) is not the most accurate I think. Though now thinking: clip the resulting value so that it doesn't exceed 4080/0xFFFF and this might be the solution.Ībout white balance. But on some type of images it fails in a spectacular way -> not safe to have it on auto. It even makes great fix for shots through glass. Take for input given image lowest 0.1% or 0.01% of pixels. I was trying also to estimate black level from the image histogram. Those ntk_leaks also elude me on black levels. I just don't find any meaningful 4080 or -4080 or any close number (also tried +/-4080/16=255) in Git2P-90 fw. It is very consistent, same size files always have black levels at the same offset. I found black levels in all amba FWs (it is 4 negative int32 instead of int16 in some newest FW). This looks to me that 800/0x3FFF is typical of Sony (like Imx117 or Imx206) while 4080/0xFFFF is typical of Panasonic (Mn34112 or mn34120). The ratios 4080/0xFFFF and 1023/0x3FFF are nearly the same! Moreover I have found black levels from Mn34112-Eken_V8S amba firmware to be -1023. It is not negligible level: 4080/0xFFFF is about the same order as 800/0x3FFF on FF8SE or Xiaomi Yi. And very nice dependence on ISO, but I would ignore that.
![fastrawviewer 1.3.3 fastrawviewer 1.3.3](https://www.fastrawviewer.com/sites/fastrawviewer.com/files/FastRawViewer_Batch-Title-Description-Setting.jpg)
Very consistent result: 4080 is the black level.
#FASTRAWVIEWER 1.3.3 ISO#
Close lens, wrap with towels, and make all iso and both short and long exposures. This is old plot of mine checking black level on Git2P-90. I wanted to check on this more, but it might be relevant to your efforts, so lets have it now.
#FASTRAWVIEWER 1.3.3 CODE#
Just in case if it is not clear: any piece of code of mine on this forum (code, plot, info, table, test image, etc) anybody can use in any way, copy/paste, commercial or not. If you compile sources maybe I could give you few changes to try to make it safer for linux or g++: it is possible that "narrowing conversion" int->dword is the culprit. Do you compile it from sources or you just wrap original exe?
![fastrawviewer 1.3.3 fastrawviewer 1.3.3](https://www.fastrawviewer.com/sites/fastrawviewer.com/files/styles/extra_large/public/FastRawViewer-1-4-4-Preferences-XMP-Use-Custom-String.png)
Usually small programs like that run fine.