Notices

I see, thanks for the write-up.
Shutter open time well makes sense in my head, but I "should" google about night-vision.

Undefined shutter open would rock, with extra low ISO you could take very interesting Long (hours!) exposure photos...

...why not buy cheap Canon (maybe A series), remove the filter there, and just put CHDK on it? Hell, i can sell you my 10 years old PowerShot G1 (just needs a new battery) which is NIR-compatible out-of-the-box! Or a second-hand years old Sony camcorder? with night mode? It's just one small box more. The N900 is not _that_ an all-rounder. At least not without frankenstein-looking. The amount of time you spend on research & modding is with very high probability more (expressed in bucks) than it would cost you to use a second-hand dedicated device (with it's own battery, btw.).

But where is the fun in that?

This is what progress is. It is good on several levels - coders get to practice and refine their skills, workarounds and unconventional fixes can sometimes lead to a new discovery and most importantly we refine our N900s.

AFAIK fCam is targeted not only at N900 so other devices would also benefit from its development.


@Estel I also made my own N900 tripod and am drooling over night photography. The sensor's 100 ISO setting is pretty well suited for low noise images.


Also this is worth looking at Post 665 @ this thread, but nobody gave it any attention (and I even made those pretty graphics)..,

thinking of experimentation
is it possible to make a motion-cam out of N900...(i think a bit tough)

but i think if we some how trigger the camera capture with the microphone
we will be able to grab the pictures/videos on any audio activity...

Well, apart of answer that You've already got from OVK and tecs, reason - in my case - is simple. I always have my N900 with me. *Always*. And you know murphy law, you can have with you (even smallest) second camera, then, after half of an year, when You won't have it for one day, a plentora of interesting photo opportunities manifest themselves. It's exactly same with heavy tripods - they're great, powerful, stable, etc, but you never have them when You really need it. So I'm using ~0,3 kg modified tripod, that suits both my photo and video needs.

When comes to NightVision device (not photos - I mean viewfinder), it's even more clear - most of the time, you aren't carrying Night vision goggles with you, yep? NightVision photography is quite different things, cause You prepare for it (most of the times) - yet, it's still super cool to have it on N900 itself.

Of course, I admit that modding-fun is great part of why the heck bother". I also agree that it would probably cost the same - minus time spent on modding - to buy ready device (yet, i must admit that I've spent almost no money on mods, except some glue, and 0,1$ parts - I'm basically self-manufacturing everything, including tripod mount itself). Well, I could say that - at least - I don't need to explain my wife why we need IR capable camera or Night Vision device

But the truth is, that modifying it, and achieving such "unbelievable" things in our N900 is a lotta hell of fun. Just ask freemangordon, why he's working on 720p video recording and viewing, while camcoders of such capabilities can be bought cheap. Damn, if we already got those mini, wideangle Tessars (that probably are too small to benefit from 5 MP anyway - fixme?), lets use it fully!

/Estel

// Edit

Should be totally trivial to write for even moderately experienced coder (which, unfortunately, I'm not). Even with trigger only on user-set dB - there are many chunks of code available for it in our extras/garage projects.

// Edit 2

Tecs, I haven't commented about Your lossless digital zoom idea, because I haven't anything of value to say about that (thus only "Thanks" button pressed) - sounds nice, but totally out of scope for my skills, to even think about way of implementing. I wonder though, if it is even possible to implements using fCam libraries (as for camera-ui(2), it's probably not, as it's using what we already got by libraries and modules, modifying GUI side).

Some time ago, I mentioned vignetting problem visible in RAWs (be it camera-ui ones, or fCAM), that was very hard to fix manually. Thanks to geohsia, we have FOSS fix for that! Look here:
http://talk.maemo.org/showpost.php?p...7&postcount=19
You can also find nice examples of what I've been talking about (cyan vignette).

If we're talking about RAW's, two questions:
1. geohsia stated, that fCAM RAW's are "better suited" to create vignette-deletion profiles (using them as reference image for CornerFix's algorithms). I've experimented with images using tripod - in exact same light conditions etc, one made with fCam, and another using camera-ui. Indeed, they're giving a little different results.

I know fCam uses different libraries, but shouldn't RAW (being sensor data dump) be close to identical?

2. Could anyone knowledgeable about N900 RAW's explain me that?
http://talk.maemo.org/showpost.php?p...2&postcount=29
Is it useless "ghost" thumbnail, remnant of JPG creation on N900 (from RAW, that we're using now), or is it indication of some useful, yet unused (maybe because some errors in creating .dng from our RAW) data left in RAW?

Hi Estel,

my camera-ui uses maemos camera library (gdigicam/gstreamer) for
creating the image. It is the camera driver which generates
the preview image and the final image.

When raw-mode is enabled, I grab the raw imagedata, create
a dng file and use the preview image from the camera driver
as the thumbnail. So, the thumbnail image is the post processed
image from the camera driver, that's the why it is looking
different than the raw image.

Taking the preview image as the thumbnail image for the raw dng
file isn't the best way. But it was simpler than writing my own
code for generating a thumbnail image. The way nokia post processed
the image data is part of the closed source components and
the parameter are undocumented or unknown.

Thanks for answer, Nicolai. If I understand it correctly, raw imagedata is the same, no matter who pick it up (fCam or camera-ui). So, fCam creates .dng file (from mentioned raw imagedata) differently? It is even possible?

Of course I mean image data itself, not thumbnail, that must be different - this case is very clear, and thanks again for explaining.

---
Fortunately, CornerFix does awesome job on (lossless) getting rid of cyan vignette, so we've overcome the biggest obstacle of getting high-quality images from RAWs. Furthermore, CornerFix is FOSS, so there is a chance that it's going to be ported to N900 (which, in tandem with ufRAW, supply us with whole 'photolab' on device).

I'm pretty sure, that other things, resulting in "better" coloration on jpeg's created by closed-source components, are just matter of proper advanced image processing in RAW developer, and with some work involved, we can create even better "general purpose" set for developing our RAW's (such profile - for example, ufRAW compatible - can be freely distributed), + still having possibility to manually tweak every image (with or without applying generic profile first), if desired.

Thanks for making mthis dream come true - without Your work on camera-ui, it wouldn't be impossible (fCam still lacks many parameter possible in camera-ui)!