Sunday, 24 June 2012

Exercise : Sensor Linear Capure

This is the first exercise of part two.... Sensor Linear Capture 


The instruction is to take any Jpeg image, I selected one taken in a recent trip to Paris.  I opened this in Photoshop and converted it to 16 bits per channel.  Image - Mode - 16 bits.




Here I have gone to image - adjustments - curves and made a curve using a few points to make it smooth.  It has made the image dark.  This is what the image would have looked like when captured, before the cameras processor has got to work.  I saved this image (Paris16bt).



I reopened the first image and put along side this one, looking from  one to the other.  I opened the histogram for both.  The dark image the histogram was pushed to the left, confirming the darkness.  The histogram for the original image was more central showing an even coverage.  I have made a note of these histograms in my notes.  What this means is that most of the levels available to represent the tones are devoted to the brightest part of the image, while the darkest parts - the shadows at the far left are actually represented by very few levels.  This also has a very important implication for noise.

I have reopened the dark image and opened the curves dialogue.  I created a curve and tried make the image look as close to the original as possible by adjusting the curve.  I did feel maybe I couldn't get it exactly, but this may be lack of photoshop knowledge.  I have the new image below (Paris16bt 2).






This has done more or less what the cameras processor does each time.  Notice the biggest effect has been on the darkest parts.  They have been lightened by what seems several stops.

If there is noise in an image it is concentrated in the shadows.  As captured, before processing in the camera, it is in a sense buried in the overall darkness.

However the strange curve that has been applied to lighten the image to a normal appearance has the unfortunate side effect of exaggerating this noise because it lightens the shadow areas so strongly.



No comments:

Post a Comment