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
|
日時: Mon Jul 21, 2014 10:52 pm 記事の件名: |
|
|
Yes, because it is a CMOS sensor, at highest gain and some gamma a pattern noise can be observed, especially if the contrast is stretched further. It may disappear when brightness is larger. It also has hot pixels.
The mean image seems calculated in an unusual way, it gives more weight to earlier frames than later frames, look at the star trails, they should be uniformly bright along the trail.
説明: |
|
ファイルサイズ: |
157.81 KB |
閲覧数: |
8527 回 |
|
_________________ Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com |
|
トップに戻る |
|
|
SonotaCo Site Admin
登録日: 2004.08.07 記事: 12671 所在地: 139.67E 35.65N
|
日時: Tue Jul 22, 2014 6:49 am 記事の件名: |
|
|
Lightningwizard wrote: | The mean image seems calculated in an unusual way, it gives more weight to earlier frames than later frames, |
This is the most important feature of UFOCapture.
It is the effect of time domain discrete low pass filter (or, simply called "exponential averaging") on each pixel brightness, that enables continuous averaging over frames.
UFOCapture uses this time domain averaging filter and space domain differential filer to detect invisible starts that may cause scintillation.
So it is easy to distinguish the actual star image and hot pixels as below.
It looks most of your masks are that on actual stars.
But anyway they are too many. I will do something for that.
説明: |
|
ファイルサイズ: |
168.13 KB |
閲覧数: |
8514 回 |
|
説明: |
|
ファイルサイズ: |
85.07 KB |
閲覧数: |
8513 回 |
|
|
|
トップに戻る |
|
|
Lightningwizard
登録日: 2007.10.19 記事: 165
|
日時: Tue Jul 22, 2014 10:25 am 記事の件名: |
|
|
SonotaCo wrote: | This is the most important feature of UFOCapture.
It is the effect of time domain discrete low pass filter (or, simply called "exponential averaging") on each pixel brightness, that enables continuous averaging over frames.
UFOCapture uses this time domain averaging filter and space domain differential filer to detect invisible starts that may cause scintillation. |
That is interesting to know
SonotaCo wrote: | It looks most of your masks are that on actual stars.
But anyway they are too many. I will do something for that. |
In my own comparison of the blue mask with the green image in ImageJ, for a case with the mask not covering the actual stars, it really masks exactly the hot pixels and not the stars. If I do an edge detect on the green channel frame, the result is highlighting of the brightest stars and not hot pixels. If I enhance the contrast and then find edges, it shows me more stars and the hot pixels.
When daylight comes I suppose it no longer detects the hot pixels because most of them were not really bright, for example 57 in RGB value, while the light of the stars is added to the background light. Maybe some glow around them is lost in the twilight, making a sharper pixel to be masked.
説明: |
Green channel is here the edge detect in ImageJ on original green channel stretched contrast 0.01%. Some blue mask pixels overlay green edges, these however are hot pixels, the stars have no blue mask. |
|
ダウンロード |
ファイル名: |
edges.png |
ファイルサイズ: |
605.07 KB |
ダウンロード回数: |
397 回 |
_________________ Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com |
|
トップに戻る |
|
|
SonotaCo Site Admin
登録日: 2004.08.07 記事: 12671 所在地: 139.67E 35.65N
|
日時: Fri Jul 25, 2014 6:36 am 記事の件名: UFOCaptureHD2 V4.10 |
|
|
UFOCaptureHD2 V4.10 that contains 4pix scintillation mask mode was released.
http://sonotaco.com/soft/download/UFOHD2_410.zip
V4.10 uses 4 pixel average for the computation of scintillation mask.
4pix mode will reduce the mask for isolated 1 pix mask for hotpixels of CMOS.
説明: |
|
ファイルサイズ: |
128.71 KB |
閲覧数: |
8441 回 |
|
説明: |
|
ファイルサイズ: |
14.63 KB |
閲覧数: |
8441 回 |
|
説明: |
|
ファイルサイズ: |
19.25 KB |
閲覧数: |
8441 回 |
|
説明: |
|
ファイルサイズ: |
17.5 KB |
閲覧数: |
8441 回 |
|
|
|
トップに戻る |
|
|
Lightningwizard
登録日: 2007.10.19 記事: 165
|
|
トップに戻る |
|
|
SonotaCo Site Admin
登録日: 2004.08.07 記事: 12671 所在地: 139.67E 35.65N
|
日時: Sat Aug 02, 2014 8:34 am 記事の件名: |
|
|
The current masking functions of UFOCapture is originally developed for SD video. The target of scintillation mask is point, not circle.
So the hot pixels of CCD is just like the target and the too fat star image on hi-res is not.
On the way of its original thinking, all star image should be covered by scintillation mask with appropriate mask size(sm size), but not slow object mask. The bigger SMsize such as 11 may cover bigger star image. But of course there is limit come from the restriction of real time processing.
Also the slow shutter time, slower than frame rate or too slow frame rateis out of design. UFOCapture watches the event duration by number of frames, but it is not measurable on those inputs.
The 10 frame of the lower limit of Head is needed for the automated analysis of meteor trajectory( to know the background level), so it will not be changed in official versions.
I developed 4pix mode on V4.10 but it seems it is not so good.
It seems some new masking technics should be invented for Point Grey.
Sorry, I went Sony 4K.
説明: |
|
ダウンロード |
ファイル名: |
z20140801_195553_TK1_A9_060s.jpg |
ファイルサイズ: |
852.19 KB |
ダウンロード回数: |
398 回 |
説明: |
|
ダウンロード |
ファイル名: |
z20140801_202235_TK1_A9_061s.jpg |
ファイルサイズ: |
686.01 KB |
ダウンロード回数: |
410 回 |
|
|
トップに戻る |
|
|
Lightningwizard
登録日: 2007.10.19 記事: 165
|
日時: Mon Aug 04, 2014 4:48 am 記事の件名: |
|
|
Oh, very nice lightning shots!
4pix mode works better than the classic mode, so I appreciate it. At larger SMsizes also masks around the small stars become large which would not be necessary. The biggest stars may need 9 pixels if computationally possible. Not sure if this would be favorable for the small ones.
I will post some images of last nights in Etc!
_________________ Weather & Photography
http://www.lightningwizard.com
EuroSprite blog:
http://eurosprite.blogspot.com |
|
トップに戻る |
|
|
tkuhmone
登録日: 2014.08.26 記事: 56
|
|
トップに戻る |
|
|
|