SonotaCo.JP
SonotaCo Network Japan Forum
SonotaCo.JP Forum Index
homeTop Page  FAQFAQ   検索検索   メンバーリストメンバーリスト   ユーザーグループユーザーグループ   登録する登録する 
 プロフィールプロフィール   プライベートメッセージをチェックするプライベートメッセージをチェックする   ログインログイン 

MinL-N

 
新しいトピックを投稿   トピックに返信    SonotaCo.JP Forum Index -> UFOCapture BBS in English
前のトピックを表示 :: 次のトピックを表示  
投稿者 メッセージ
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Mon Nov 09, 2009 6:25 am    記事の件名: MinL-N 引用付きで返信

Hi SonotaCo!

I'm capturing a lot of sprites and elves again lately, also now as I type over winter storms north of Spain.
I am still using sub-Detect Level Noise Tracking settings in order to get the best detection for very weakly luminous events possible, combined with capture card sharpening turned off. For example, my Detect Level Noise Tracking with MinL-N of 0 suggests 20-21. But I can go below that to 18-19 without problems. What I'd like to be possible is to use Detect Level Noise Tracking allowing negative values of MinL-N. The advantage is that it would still respond to changes in noise conditions. For example noise may increase when the daylight arrives, and if I let it at the fixed hyper-sensitive setting it could start recording very long clips.

greetings,
Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Mon Nov 09, 2009 8:32 am    記事の件名: Re: MinL-N 引用付きで返信

Hi Oscar
It seems your skill improves further.
I saw your recent still color images and new clips. That's wonderful and beautiful.
Lightningwizard wrote:
What I'd like to be possible is to use Detect Level Noise Tracking allowing negative values of MinL-N.

Negative MinL-N may be useful for faint but large events.
OK, I will try it. Wait for a while.
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Mon Nov 09, 2009 10:57 am    記事の件名: Re: MinL-N 引用付きで返信

The package below contains UFOCaptureV2 V2.23 which allows negative MINL-N.
It was not tested enough. So the negative MinL-N setting with DLratio < 100
may cause some troubles depends on the scene (such as movement of the clouds).
Therefore I publish it here as a trial version for the time being.

enjoy!



UFO223.zip
 説明:
UFOCaptureV2 V2 23 trial

ダウンロード
 ファイル名:  UFO223.zip
 ファイルサイズ:  2.89 MB
 ダウンロード回数:  534 回

トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Tue Nov 10, 2009 4:21 am    記事の件名: 引用付きで返信

Thanks!!
It seems useful, there are of course issues because of the increased sensitivity.
I set Detect level to <100% and MinL-N to -2 and detected already several elves right from the start of the evening, near Corsica, so it is a good case to test.
MinL-N of -2 seems relatively safe, as soon as you go down step by step you hit a strong increase usually at -3. It then helps to turn on Dark Object Mask (level 0). I set a size of 5 and it sometimes detects something not related to lightning, such as planes and subtle lens reflections caused by lights from cars and subtle flickering of streetlights. But that is what you get if you want sensitive! In such cases it is hard to tell what triggered the camera, the blue pixels are scattered without clustering.
I further decreased MinL-N to -4, and I get even very weak elves I would hardly see with my eyes on the screen. I did have to set size to 40 (with Dark Object Mask at 0)! Slow object mask seems to work, but flashes of airplanes will have a much larger radius and will become more annoying....

UPDATE: one tricky thing to consider is that the noise level tracking can change by 1 unit, so if I run with MinL-N at -4 and the 40 as size, and the auto detect level jumps by 1, it may require a lower Detect Size to retain sensitivity... I'm not sure how important the size is since the object to be detected should be large anyway. In that case, one may have to try Detect Noise Size Tracking in addition to Level. But if it fluctuates too much it can be a problem.
I let you know if I discover more.

By the way, do Japanese observers get many elves? I do not see them on the sprite part of the forum? It looks like there must be an awful lot of elves there over winter storms. In many of them I can see gravity wave patterns, very interesting Smile

Oscar



M20091109_190602__LW.avi
 説明:
Weakest elve detected! MinL-N -4, Detect Size 40, DOM 0
Usually such events (but often brighter) are noticed on screen but not detected. This one is at the limit of my own detection. Warning.... unspectacular video ;) XviD format.

ダウンロード
 ファイル名:  M20091109_190602__LW.avi
 ファイルサイズ:  3.72 MB
 ダウンロード回数:  553 回


_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Tue Nov 10, 2009 10:01 am    記事の件名: 引用付きで返信

Thank you for reporting. I am happy to know this update seems useful.

In Japan we notice elves only on a few to 10% of sprite events.
But it might depend our sensitivity or transparency of the sky.
We may have lost many weak ones like your weakest sample.

Gravity wave pattern Exclamation .... Oh, very interesting.... I have not recognised it.
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Tue Nov 10, 2009 1:24 pm    記事の件名: 引用付きで返信

It works okay generally, but for unattended use I would not recommend MinL-N of -3 or lower. This is in the region of the noise spectrum where the number of pixels with the corresponding brightness change grows rapidly with decreasing MinL-N. So an auto level change of 1 could result in a varying detection efficiency. Once you see the Size bar display constant activity, it is difficult to keep it stable. For example, it could fluctuate between 10 and 20, with occasional jumps to 30. But setting Size above 30 to avoid triggering would probably mean that small objects are not going to be detected.

The slow object mask can act normally, but occasionally it hunts noise somewhere. I think that is not desirable, as it may cancel an event as I found out last year. The mask behaves often quite erratic also when tracking planes, probably more so at more sensitive detection settings. I am thinking of ways how it could be improved (not so easy)
For meteors MinL-N = 0 is fine.

Other ideas:

1) Instant trigger using a pre-trigger buffer and a hotkey (or spacebar when window active) to be sure an event you see on screen is recorded, also when in Preview. It would only record the specified number of frames before and after this. The practical issue would be that you need a quite long pre-trigger buffer like 30-60 frames which you would not prefer for standard detection.

2) An option to create a time lapse series of images, using integration over several seconds like an astro camera to obtain the best image of the starry sky. Perhaps images need to be converted to linear (gamma 1.0) first before stacking? I'm curious what that would look like... Smile

Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Tue Nov 10, 2009 6:01 pm    記事の件名: 引用付きで返信

Lightningwizard wrote:
1) Instant trigger using a pre-trigger buffer and a hotkey (or spacebar when window active) to be sure an event you see on screen is recorded, also when in Preview. It would only record the specified number of frames before and after this. The practical issue would be that you need a quite long pre-trigger buffer like 30-60 frames which you would not prefer for standard detection.

Well, how is clicking twice the "Rec" button of Live sheet. Though it is not a key action, it can cause manual trigger.
Or, how is shouting to a microphone that is connected to pc(using audio trigger of UFOCaptureV2) Wink .
I am connectomg a pair of loop antennas with microphone amplifier to one of my PC. It records many VLF events with the sky images.
...and buffering of 30-60 frames is usual for meteor observers to record the faint scene of starting part.

Lightningwizard wrote:
2) An option to create a time lapse series of images, using integration over several seconds like an astro camera to obtain the best image of the starry sky. Perhaps images need to be converted to linear (gamma 1.0) first before stacking? I'm curious what that would look like... Smile

Did you see the green channel of M*M.bmp ?
Though it is not series and it is averaged image using exponential averaging on time.(You can change the time constant by setting of SMspeed of scintillation mask.)....
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Wed Nov 11, 2009 12:54 am    記事の件名: 引用付きで返信

SonotaCo wrote:

Well, how is clicking twice the "Rec" button of Live sheet. Though it is not a key action, it can cause manual trigger.
Or, how is shouting to a microphone that is connected to pc(using audio trigger of UFOCaptureV2) Wink .

That's actually a good idea! It will also record the shouting in the movie Very Happy

SonotaCo wrote:

I am connectomg a pair of loop antennas with microphone amplifier to one of my PC. It records many VLF events with the sky images.
...and buffering of 30-60 frames is usual for meteor observers to record the faint scene of starting part.

Yes, I often used an AM radio (or two!). Now I should replace my audio cable again, for some reason they fail all the time... I also tried using it for detection, but it was not so succesful for sprites, it sometimes did not trigger on the sferics and the signal should be free of noises. I should get more experience with this function.

Usually I use only 10 frames leading time, so for best reaction I better increase that Wink The reason is smaller files and since I use Vista it can take longer for the detection to resume, it improved in SP1, but still varies between 0.5-4 seconds for me. I don't know if this is any better in Windows 7, but the new DirectX 11 does not have effect.

SonotaCo wrote:

Did you see the green channel of M*M.bmp ?
Though it is not series and it is averaged image using exponential averaging on time.(You can change the time constant by setting of SMspeed of scintillation mask.)....


Yes, I checked that occasionally with the ImageAnalyzer software. Looks like I should use this, exponential averaging I assume is what I want. It may be better than RegiStax averaging. The images of 2008 I've put on my website use a stacked image with the sprite frames merged into this, in order to get the best of star background, sprite and low noise (I still keep it interlaced, removing that creates other aesthetic issues)
Thanks for your suggestions!

Oscar

_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Wed Nov 11, 2009 1:39 am    記事の件名: 引用付きで返信

Lightningwizard wrote:

Yes, I checked that occasionally with the ImageAnalyzer software. Looks like I should use this, exponential averaging I assume is what I want. It may be better than RegiStax averaging.


ah, it was the reverse.... I started to use RegiStax for this because after adjusting the contrast of the green channel of the M images, it shows vertical lines of the capture device (Dazzle) too obviously, and grey shade banding (8 bits effect?) Sad. Does your Canopus firewire card show any vertical banding?
I was hoping to see any structure in the background sky, like the Milky Way or airglow patterns. I noticed those in images from someone in Switzerland using a camera with internal integration like a Mintron, or someone in Texas with a Watec 120N/Astrovid Stellacam 2. Perhaps there is no software substitute for this or I need to have a darker observation site....

Oscar
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Wed Nov 11, 2009 9:35 am    記事の件名: 引用付きで返信

Lightningwizard wrote:
it shows vertical lines of the capture device (Dazzle) too obviously, and grey shade banding (8 bits effect?) Sad.

It is hard to imagine Confused Let me see a sample.
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Wed Nov 11, 2009 11:28 am    記事の件名: 引用付きで返信

I attach an example. I split the RGB with Image Analyzer. Then you see it is quite grey, so I set a black point, here that would be on the dark part of the roof. The banding from the capture becomes apparent but also the greyscale banding.

I did a comparison with an image coming out of Registax from the movie, which should be shorter than the period the M was calculated and therefore more rough. That one comes out with a better black point which can still be improved. But when tuning black, white and gamma settings, this image shows less obvious banding. I think it is simply because of the smoothness of the final image that makes a more dithered impression for the RegiStax version and banded result of the M version. So my problem is mostly my Dazzle device and 8-bit adjustments, not the M result.
The visibility of stars is quite the same between the two.

Oscar

EDIT: wait - the M result can be better by converting to 24 bit or float! It seems mostly user error Embarassed



stack_vs_M_processed.png
 説明:
example after black point. Left: Registax from video, Right: M green channel
 ファイルサイズ:  363.53 KB
 閲覧数:  9621 回

stack_vs_M_processed.png



_________________
Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com
トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Wed Nov 11, 2009 1:13 pm    記事の件名: 引用付きで返信

the M result looks better by converting to 24 bit, then add some white noise, levels, then paste in Paint.NET, add light Surface Blur and sharpening.
Now I only need a better capture card Wink



M_processed_better.png
 説明:
just a test image
 ファイルサイズ:  586.46 KB
 閲覧数:  10613 回

M_processed_better.png


トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Wed Nov 11, 2009 2:49 pm    記事の件名: 引用付きで返信

Hi Oscar, thank you for samples.
Now I understood what you saying. Oh, it is very weak vertical line noise around LSB.....

Though I am not sure, I think this is analog noise which is added before the digitizer.
Because once I have seen similar noise on analog monitor which directory connected to camera.
It may be caused by camera itself , power supply, or impedance mismatch of cable or connectors.
There is similar noise on some of my system. So... it might be very difficult problem.



stack_vs_m_processed_2.png
 説明:
the first sample
clear vertical lines....
 ファイルサイズ:  105.51 KB
 閲覧数:  10604 回

stack_vs_m_processed_2.png



m_processed_better_G.png
 説明:
second sample looks better but ...
 ファイルサイズ:  641.1 KB
 閲覧数:  10604 回

m_processed_better_G.png



M20091108_034730_TK1_S6MG.png
 説明:
My S6 system ... has little same noise
 ファイルサイズ:  87.87 KB
 閲覧数:  10604 回

M20091108_034730_TK1_S6MG.png



M20091104_003321_TK1_S2MG2.png
 説明:
My S2 system.... almost none
 ファイルサイズ:  103.15 KB
 閲覧数:  10604 回

M20091104_003321_TK1_S2MG2.png


トップに戻る
ユーザーのプロフィールを表示  
Lightningwizard



登録日: 2007.10.19
記事: 165

記事日時: Wed Nov 11, 2009 10:40 pm    記事の件名: 引用付きで返信

Thanks for showing yours! Too bad banding is even there with the Canopus which should be top-notch quality. I've seen on the web some comparisons between capture devices and the degree of banding varies. Electronic interference noise I get sometimes as well, those patterns are diagonal.
The banding could perhaps be reduced. The M image is 8 bits per color channel, could this be increased? It would increase the file size and for the other channels it is not as useful, but...
One could perhaps save the stacked field in 16 bit greytones as PNG lossless compressed?
Even if I convert the M bmp to png lossless it becomes only 82 kB which sounds good. I don't know about encoding/decoding performance.
トップに戻る
ユーザーのプロフィールを表示  
SonotaCo
Site Admin


登録日: 2004.08.07
記事: 12653
所在地: 139.67E 35.65N

記事日時: Thu Nov 12, 2009 9:55 am    記事の件名: 引用付きで返信

Lightningwizard wrote:
One could perhaps save the stacked field in 16 bit greytones as PNG lossless compressed?
Even if I convert the M bmp to png lossless it becomes only 82 kB which sounds good. I don't know about encoding/decoding performance.

16 bit png! .... I never thought of that. ..... interesting Confused
トップに戻る
ユーザーのプロフィールを表示  
特定期間内の記事を表示:   
新しいトピックを投稿   トピックに返信    SonotaCo.JP Forum Index -> UFOCapture BBS in English All times are GMT + 9 Hours
Page 1 of 1

 
移動先:  
新規投稿: 不可
返信投稿: 不可
記事編集: 不可
記事削除: 不可
投票参加: 不可
このフォーラムで添付ファイルを投稿 できません
このフォーラムでファイルをダウンロード できます


Powered by phpBB © 2001, 2005 phpBB Group
Copyright ©2004 SonotaCo Network. All Rights Reserved.