Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: -Andrew- on November 29, 2005, 10:52:28 pm
-
[span style=\'font-size:8pt;line-height:100%\']Can someone spare the time to give me a brief explanation of the process of interpolation? Or at least refer me to some good references?[/span]
-
I think the guy that does DCRaw has some articles about it. Or at least he references some papers on methods of interpolation.
In non fovean digital cameras each pixel only has information about one color. Not all 3. Interpolation tries to figure out what the other two colors are based on infromation in the adjoining pixels. A simple method is to just average the colors in the adjacent pixels.
-
Dark: you are quite correct that hue interpolation is used in most cameras to turn monochrome sensor data into RGB. In addition, interpolation is one name for the process also referred to as up-sampling, down-sampling, up-rezzing, down-rezzing, etc.
Let's say I have a 2000x3000 pixel image from my camera and want to turn that into a 4000x6000 pixel image for printing. Think of a sheet of graph paper (a page divided into small squares). I colour square 1 (upper left) with the colour recorded by pixel 1 from the image. I then skip square 2 (just right of square 1) and colour square 3 with the colour recorded by pixel 2 from the image. And so on. When I'm done I have stretched the image to cover the new 4000x6000 area but have left holes all over. Interpolation is the formula used to guess the RGB values to colour the empty squares with based on the colour of their neighbours.
If this is unclear I can create an illustration later today.
-
Here's a short explanation (http://en.wikipedia.org/wiki/Bayer_filter)
-
Here's a very mathematical description comparing some of the more common interpolation methods.
Algorithm's for Bayer Interpolation (http://graphics.cs.msu.su/en/publications/text/prog2004lk.pdf)
-
Dark: you are quite correct that hue interpolation is used in most cameras to turn monochrome sensor data into RGB. In addition, interpolation is one name for the process also referred to as up-sampling, down-sampling, up-rezzing, down-rezzing, etc.
Let's say I have a 2000x3000 pixel image from my camera and want to turn that into a 4000x6000 pixel image for printing. Think of a sheet of graph paper (a page divided into small squares). I colour square 1 (upper left) with the colour recorded by pixel 1 from the image. I then skip square 2 (just right of square 1) and colour square 3 with the colour recorded by pixel 2 from the image. And so on. When I'm done I have stretched the image to cover the new 4000x6000 area but have left holes all over. Interpolation is the formula used to guess the RGB values to colour the empty squares with based on the colour of their neighbours.
If this is unclear I can create an illustration later today.
[a href=\"index.php?act=findpost&pid=52480\"][{POST_SNAPBACK}][/a]
Dale: How are you doing? A long time since I have spoken to you or seen your posts here. Your posts and explanations are as clear and informative as ever. Sharon and I are doing great and retired year around to LBI.
Very Best...
Dave Gurtcheff