How to do FF calibration in PS?

LT
夸克星
文章: 2507
註冊時間: 週日 17 8月, 2003 23:20
來自: YL
聯繫:

How to do FF calibration in PS?

文章 LT » 週二 15 11月, 2011 13:12

Hi,

I have very little knowledge in imaging processing and their theories behind. So, if that is a silly question, please don't laugh at me, and I will be most appreciated for any input from you.

Just as titled. How to do it? There is only Subtract or Difference in application/calculation. I had tried, but the field was not flat calibrated. From books, they tell the calibration should be by dividing. Is it true? Is there such an operation in PS. How to "call" it out? My PS version is CS2.

I am a stupid man. So, I would be most grateful if any printscreen illustration attached.

Thanks in advance.

LT

頭像
XR250
夸克星
文章: 6513
註冊時間: 週一 13 12月, 2004 09:43
來自: NT

文章 XR250 » 週二 15 11月, 2011 15:05

我也是用Photoshop CS2,CS2是沒有divide的,只可選擇Subtract

頭像
鄧登凳
夸克星
文章: 9493
註冊時間: 週一 03 8月, 2009 17:15
來自: 3rd planet of solar system

文章 鄧登凳 » 週二 15 11月, 2011 15:53

In flat field calibration, the light frame should be adjusted according to the proportion of light received in the flat field. So ratio mathematics should be use. It is basically division.

Photoshop (I suspect that it is for most versions, although I have only used CS4 and element7) has a function/blending mode call "color dodge". The maths (assuming 8bit colour 0-255) is division:

R = Min(bottom * 255 /(255-top), 255) , if top=255, R=255
bottom = RGB color value of the base layer
top = RGB color value of the adjustment layer
Min = minimum function to ensure the value is not over 255
R = resulting pixel color value

Now, since the invert (255-top) is used, the FF has to undergone an inversion first (black to white and white to black) first. Then use color dodge to combine the inverted FF (top) with the LF (bottom) through the division mathematics.

Sometimes, it is necessary to lighten the inverted FF so as to avoid saturating the product.

PS:
1. Blending mode maths: http://www.vanderlee.com/tut_fm_mixingmodes.html

2. I always find the names of photoshop blending modes bewildering, so don't ask me why dividing with the inverted top is called "color dodge".

3. I do not have photoshop in my office machine. So cannot do screen capturing here. However, inversion is one of the basic adjustment functions and color dodge is one of the blending modes (as the difference blending mode). Both should not be difficult to find.

4. FF in DSS should be more accurate since the division is applied to individual subframes before combining.

頭像
XR250
夸克星
文章: 6513
註冊時間: 週一 13 12月, 2004 09:43
來自: NT

文章 XR250 » 週二 15 11月, 2011 18:50

我剛剛試了一次用了invert了的flatfield+Color dodge來處理FF。結果能夠做到divide flat field的效果。

但我想問Color dodge原本是用來減(除?)低color contrast,用一張invert了的Flat field來進行Color dodge會唔會連color也變flat了?

另外"FF in DSS should be more accurate since the division is applied to individual subframes before combining.", 我覺得Deepskystacker做得更好,正確說法是DSS直接對RAW進行divide,而不是"除"了之後才把subframe合成。

要知道RAW經過debayering之後,已經變了一張圖片(JPG/BMP/TIF/FIT)。雖然對圖片進行divide FF還是可行的,但要先確定圖片是liner地從RAW轉換出來,才能用另一張liner轉換出來的FF來"除"。所以在PS上處理FF,並不是最好的方法,但肯定比較容易上手。

其實我通常都是在PS上處理,原因是我沒有拍FF的習慣,大多是事後用PixInsight LE生出來,然後在PS上"減"FF;講到尾都係我~~~懶。 :) :)

PixInsight LE是有divide FF的功能,LT可以研究一下。

頭像
Subaru
夸克星
文章: 6693
註冊時間: 週三 02 7月, 2003 19:27

文章 Subaru » 週二 15 11月, 2011 23:18

跟據理論:
Intensity = (Light - Dark - Bias) / (Flat - Dark - Bias)

做image calibration, flat是要用除做operant. 用減的方法去做, 個D-range會低咗.

其實好多天文專用軟件如Deepskystacker, IRIS或Nebulosity已經係利用以上formula去做calibration.

如其用Photoshop去攪, 不如拍好d flat, bias及dark, 用呢d天文專用軟件(DSS及IRIS是免費的!)直接做calibration, 效果會比用PS攪一大輪來得直接.

用Pixinsight做background subtraction, 已經係做image calibration之後的一個step, 來減少gradian. Image calibration係first step, 如果比較認真地去攪係無得慳的.

頭像
鄧登凳
夸克星
文章: 9493
註冊時間: 週一 03 8月, 2009 17:15
來自: 3rd planet of solar system

文章 鄧登凳 » 週二 15 11月, 2011 23:26

XR250 寫:我剛剛試了一次用了invert了的flatfield+Color dodge來處理FF。結果能夠做到divide flat field的效果。

但我想問Color dodge原本是用來減(除?)低color contrast,用一張invert了的Flat field來進行Color dodge會唔會連color也變flat了?

另外"FF in DSS should be more accurate since the division is applied to individual subframes before combining.", 我覺得Deepskystacker做得更好,正確說法是DSS直接對RAW進行divide,而不是"除"了之後才把subframe合成。

要知道RAW經過debayering之後,已經變了一張圖片(JPG/BMP/TIF/FIT)。雖然對圖片進行divide FF還是可行的,但要先確定圖片是liner地從RAW轉換出來,才能用另一張liner轉換出來的FF來"除"。所以在PS上處理FF,並不是最好的方法,但肯定比較容易上手。

其實我通常都是在PS上處理,原因是我沒有拍FF的習慣,大多是事後用PixInsight LE生出來,然後在PS上"減"FF;講到尾都係我~~~懶。 :) :)

PixInsight LE是有divide FF的功能,LT可以研究一下。
正如上面所說, PS那些「混合選項」(blending modes)旳名字使人莫名其妙, 用反相(invert)做除數叫「顏色減淡」(color dodge), 當然分母是正數的整數, 除完的商一定比原本小啦, 那為何不叫比例減光而叫顏色減淡, 我就不懂了, 那會不會顏色真的變後很淡, 我的有限經驗中是不會的, 而即使略有減淡, 都很易用飽和度(saturation)救回來。

在PS中如用TIFF, 應是沒有損失數據的, 只要LF和FF都用同一模式產生TIFF, 應不會有問題, 反而是拍攝時慢速產生的追蹤誤差或做了drizzle, 那疊完的圖和平場很可能不是完全對位的, 那就會有誤差。DSS做平場, 不會受這個影響。

頭像
noodle
夸克星
文章: 2835
註冊時間: 週五 04 7月, 2003 16:34
來自: 地球
聯繫:

文章 noodle » 週三 16 11月, 2011 07:25

Subaru 寫:跟據理論:
Intensity = (Light - Dark - Bias) / (Flat - Dark - Bias)

做image calibration, flat是要用除做operant. 用減的方法去做, 個D-range會低咗.

其實好多天文專用軟件如Deepskystacker, IRIS或Nebulosity已經係利用以上formula去做calibration.

如其用Photoshop去攪, 不如拍好d flat, bias及dark, 用呢d天文專用軟件(DSS及IRIS是免費的!)直接做calibration, 效果會比用PS攪一大輪來得直接.

用Pixinsight做background subtraction, 已經係做image calibration之後的一個step, 來減少gradian. Image calibration係first step, 如果比較認真地去攪係無得慳的.
我都覺得用PS做calibration無著數,正如Subaru所講用DSS等軟件來得方便、直接。:D
用PS整vignetting、gradient就有好多方法,有時間可以睇睇:
http://www.astropix.com/HTML/J_DIGIT/VIGNET.HTM
http://starmatt.com/articles/av.html
http://www.flemingastrophotography.com/gradients.html

頭像
noodle
夸克星
文章: 2835
註冊時間: 週五 04 7月, 2003 16:34
來自: 地球
聯繫:

文章 noodle » 週三 16 11月, 2011 07:39

鄧登凳 寫:
XR250 寫:我剛剛試了一次用了invert了的flatfield+Color dodge來處理FF。結果能夠做到divide flat field的效果。

但我想問Color dodge原本是用來減(除?)低color contrast,用一張invert了的Flat field來進行Color dodge會唔會連color也變flat了?

另外"FF in DSS should be more accurate since the division is applied to individual subframes before combining.", 我覺得Deepskystacker做得更好,正確說法是DSS直接對RAW進行divide,而不是"除"了之後才把subframe合成。

要知道RAW經過debayering之後,已經變了一張圖片(JPG/BMP/TIF/FIT)。雖然對圖片進行divide FF還是可行的,但要先確定圖片是liner地從RAW轉換出來,才能用另一張liner轉換出來的FF來"除"。所以在PS上處理FF,並不是最好的方法,但肯定比較容易上手。

其實我通常都是在PS上處理,原因是我沒有拍FF的習慣,大多是事後用PixInsight LE生出來,然後在PS上"減"FF;講到尾都係我~~~懶。 :) :)

PixInsight LE是有divide FF的功能,LT可以研究一下。
正如上面所說, PS那些「混合選項」(blending modes)旳名字使人莫名其妙, 用反相(invert)做除數叫「顏色減淡」(color dodge), 當然分母是正數的整數, 除完的商一定比原本小啦, 那為何不叫比例減光而叫顏色減淡, 我就不懂了, 那會不會顏色真的變後很淡, 我的有限經驗中是不會的, 而即使略有減淡, 都很易用飽和度(saturation)救回來。

在PS中如用TIFF, 應是沒有損失數據的, 只要LF和FF都用同一模式產生TIFF, 應不會有問題, 反而是拍攝時慢速產生的追蹤誤差或做了drizzle, 那疊完的圖和平場很可能不是完全對位的, 那就會有誤差。DSS做平場, 不會受這個影響。
用tiff無數據損失,不過raw轉tiff會唔會stretch左嫁?
stretch左先做FF,有d問題喎? [鬼點子]
tracking error、drizzling對FF好似無影響?
我估無人drizzle完先去calibrate掛? [微笑]

頭像
XR250
夸克星
文章: 6513
註冊時間: 週一 13 12月, 2004 09:43
來自: NT

文章 XR250 » 週三 16 11月, 2011 11:25

noodle 寫:
鄧登凳 寫:
正如上面所說, PS那些「混合選項」(blending modes)旳名字使人莫名其妙, 用反相(invert)做除數叫「顏色減淡」(color dodge), 當然分母是正數的整數, 除完的商一定比原本小啦, 那為何不叫比例減光而叫顏色減淡, 我就不懂了, 那會不會顏色真的變後很淡, 我的有限經驗中是不會的, 而即使略有減淡, 都很易用飽和度(saturation)救回來。

在PS中如用TIFF, 應是沒有損失數據的, 只要LF和FF都用同一模式產生TIFF, 應不會有問題, 反而是拍攝時慢速產生的追蹤誤差或做了drizzle, 那疊完的圖和平場很可能不是完全對位的, 那就會有誤差。DSS做平場, 不會受這個影響。
用tiff無數據損失,不過raw轉tiff會唔會stretch左嫁?
stretch左先做FF,有d問題喎? [鬼點子]
tracking error、drizzling對FF好似無影響?
我估無人drizzle完先去calibrate掛? [微笑]

係,TIFF只是其中一個無數據損失的容器(file format),RAW才是這個"容器"要裝的DATA。正如Subaru所說"要認真的地去攪",要處理的對像應該是RAW,並不是16bit或32bit的TIFF/FIT/....。

題外話,DSLR的RAW真是RAW嗎?最少nikon以前的RAW並不是真正的RAW,要nikon mode3這方法才是比較接近RAW的DATA。Canon/Nikon/Pentax/Sony等等的RAW有幾RAW, 大家有資料嗎?

LT
夸克星
文章: 2507
註冊時間: 週日 17 8月, 2003 23:20
來自: YL
聯繫:

文章 LT » 週三 16 11月, 2011 13:33

Thanks in advance for your input.

I think it is correct that each RAW frame will be FF, DF, Bias calibrated before stacking in DSS. So, up to now with those image processing softwares I know, no one can do FF calibration the best as DSS. But if DSS does the stacking with RAW, the resultant stacked frame seems to have "lots of data" lost (colour depth and brightness range), and also generate lots of very strange noise-like signal to the background. So, I always found it is very difficult to be corrected to my "standard". However, when the RAW is converted into TIFF and then stacked in DSS, the brightness range and color depth make a very big difference. The details and colour can be corrected to be very pleasing to me in PS. However, the vignetting is still there. I had tried to convert a RAW FF into TIFF FF, and then let DSS do the FF calibration, but failed. So, I raise such a question.

Why the stacked image in DSS loses the so-called "lots of data"? Is it just my own problem? Or do I not have correct stacking parameter settings in DSS?

I will attach some cropped images later to illustrate my problem with DSS, and see if anyone can help me to fix it up.

LT

頭像
XR250
夸克星
文章: 6513
註冊時間: 週一 13 12月, 2004 09:43
來自: NT

文章 XR250 » 週三 16 11月, 2011 16:48

LT 寫:Thanks in advance for your input.

I think it is correct that each RAW frame will be FF, DF, Bias calibrated before stacking in DSS. So, up to now with those image processing softwares I know, no one can do FF calibration the best as DSS. But if DSS does the stacking with RAW, the resultant stacked frame seems to have "lots of data" lost (colour depth and brightness range), and also generate lots of very strange noise-like signal to the background. So, I always found it is very difficult to be corrected to my "standard". However, when the RAW is converted into TIFF and then stacked in DSS, the brightness range and color depth make a very big difference. The details and colour can be corrected to be very pleasing to me in PS. However, the vignetting is still there. I had tried to convert a RAW FF into TIFF FF, and then let DSS do the FF calibration, but failed. So, I raise such a question.

Why the stacked image in DSS loses the so-called "lots of data"? Is it just my own problem? Or do I not have correct stacking parameter settings in DSS?

I will attach some cropped images later to illustrate my problem with DSS, and see if anyone can help me to fix it up.

LT

"tried to convert a RAW FF into TIFF FF, and then let DSS do the FF calibration, but failed. So, I raise such a question."
這就是我跟Noodle所講,RAW>TIFF 時並不一定是linear地convert,而是有機會nonlinear或stretch了的,然後用這個nonlinear地convert出來的FF去divide原圖,就會出現問題。

所以LT原本的方法FF, DF, Bias (RAW) 都一拼交給DSS處理是正確的。問題只是FF是否真的拍得理想。個人經驗拍一堆理想的FF,比拍天文相更難

LT
夸克星
文章: 2507
註冊時間: 週日 17 8月, 2003 23:20
來自: YL
聯繫:

文章 LT » 週三 16 11月, 2011 17:07

XR250 寫:這就是我跟Noodle所講,RAW>TIFF 時並不一定是linear地convert,而是有機會nonlinear或stretch了的,然後用這個nonlinear地convert出來的FF去divide原圖,就會出現問題。
I agree it should be.

XR250 寫:所以LT原本的方法FF, DF, Bias (RAW) 都一拼交給DSS處理是正確的。問題只是FF是否真的拍得理想。個人經驗拍一堆理想的FF,比拍天文相更難
I also think about it . Then what / how is a good/excellent FF? Correct amount of exposure = ? (i.e. where should the histogram be located ?)
Some said it should be about 1/5 from left ? Is it right?

Noodle,
I think you have read a lot of articles from internet. Could you have more input /recommendation on this issue?

Thanks.

LT

頭像
noodle
夸克星
文章: 2835
註冊時間: 週五 04 7月, 2003 16:34
來自: 地球
聯繫:

文章 noodle » 週三 16 11月, 2011 17:20

XR250 寫:
LT 寫:Thanks in advance for your input.

I think it is correct that each RAW frame will be FF, DF, Bias calibrated before stacking in DSS. So, up to now with those image processing softwares I know, no one can do FF calibration the best as DSS. But if DSS does the stacking with RAW, the resultant stacked frame seems to have "lots of data" lost (colour depth and brightness range), and also generate lots of very strange noise-like signal to the background. So, I always found it is very difficult to be corrected to my "standard". However, when the RAW is converted into TIFF and then stacked in DSS, the brightness range and color depth make a very big difference. The details and colour can be corrected to be very pleasing to me in PS. However, the vignetting is still there. I had tried to convert a RAW FF into TIFF FF, and then let DSS do the FF calibration, but failed. So, I raise such a question.

Why the stacked image in DSS loses the so-called "lots of data"? Is it just my own problem? Or do I not have correct stacking parameter settings in DSS?

I will attach some cropped images later to illustrate my problem with DSS, and see if anyone can help me to fix it up.

LT

"tried to convert a RAW FF into TIFF FF, and then let DSS do the FF calibration, but failed. So, I raise such a question."
這就是我跟Noodle所講,RAW>TIFF 時並不一定是linear地convert,而是有機會nonlinear或stretch了的,然後用這個nonlinear地convert出來的FF去divide原圖,就會出現問題。

所以LT原本的方法FF, DF, Bias (RAW) 都一拼交給DSS處理是正確的。問題只是FF是否真的拍得理想。個人經驗拍一堆理想的FF,比拍天文相更難
一個好"FF"真系唔易拍 :wink:
拍唔到點算?
呢個時你需要:

http://www.rc-astro.com/resources/Gradi ... /index.php

呢個PS plugin容易用、效果唔錯,$49.9唔省得 :D

最正就系:
http://pixinsight.com/
可以睇睇個tutorial..
http://pixinsight.com/videos/DBE_M42_Example/en.html

貴d不過好多工具,加加埋埋都算抵買 [Good Job]
不過唔易用,當年系農莊睇完Paul少示範,即時中毒 [Good Job]

頭像
noodle
夸克星
文章: 2835
註冊時間: 週五 04 7月, 2003 16:34
來自: 地球
聯繫:

文章 noodle » 週三 16 11月, 2011 17:25

LT 寫:
XR250 寫:這就是我跟Noodle所講,RAW>TIFF 時並不一定是linear地convert,而是有機會nonlinear或stretch了的,然後用這個nonlinear地convert出來的FF去divide原圖,就會出現問題。
I agree it should be.

XR250 寫:所以LT原本的方法FF, DF, Bias (RAW) 都一拼交給DSS處理是正確的。問題只是FF是否真的拍得理想。個人經驗拍一堆理想的FF,比拍天文相更難
I also think about it . Then what / how is a good/excellent FF? Correct amount of exposure = ? (i.e. where should the histogram be located ?)
Some said it should be about 1/5 from left ? Is it right?

Noodle,
I think you have read a lot of articles from internet. Could you have more input /recommendation on this issue?

Thanks.

LT
知幾多寫幾多啦,都寫得7788啦 :wink:
唔識、唔清楚個d資料,唔敢扮高手亂咁"老點"LT兄 [微笑]

頭像
鄧登凳
夸克星
文章: 9493
註冊時間: 週一 03 8月, 2009 17:15
來自: 3rd planet of solar system

文章 鄧登凳 » 週三 16 11月, 2011 18:32

noodle 寫: 用tiff無數據損失,不過raw轉tiff會唔會stretch左嫁?
stretch左先做FF,有d問題喎? [鬼點子]
tracking error、drizzling對FF好似無影響?
我估無人drizzle完先去calibrate掛? [微笑]
由RAW轉TIFF有很多不同的軟件, 很難說有沒有stretch, 不過如果stretching是線性的, 掂解會有問題??

有Tracking error, 疊完的LF的像素, 包括在感光晶片上不同位置像素的收集的光度結合而成, 那不論減或除任何單一像素的數據, 都是不正確的。

Drizzle一是在拍攝時由導星軟件配合拍攝軟件做, 一是在疊相的時候做, 拍完疊完才去PS, 那在用PS去做FF calibrate時, 不是"drizzle完先去calibrate"又算是什麼?

回覆文章

回到「天文圖像處理技巧 Astro Image Processing Technique」

誰在線上

正在瀏覽這個版面的使用者:沒有註冊會員 和 5 位訪客