Saturday, 26 May 2012

Exercise: Sensor Linear Capture

This project and exercise discusses some of the processing performed by digital cameras before it is presented to the user. A gamma correction curve is applied to bring a linear captured image to a normal appearance. This exercise simulates what happens in the camera, and draws our attention to the fact as a result of the in-camera processing, noise in the shadows/dark areas will be increased the most.

Starting with my original image (below), I then applied a curve to simulate what is initially captured in the camera (below)
Original image from camera
Gamma curve applied to simulate what is really 'raw' in the camera

The resultant image is below (note histogram in top corner, squeezed to the low end). The image looks very dark, highly constrasty and with some colour issues (e.g. blue leaves). In the dark areas there appears to be almost no detail at all.
Gamma curve applied to image
I now applied another curve to the edited image to try to bring it back to the 'original'. The resultant image is below (note that it's very difficult to bring it back to the original, as the curve editing is quite subtle and small changes have big effects when all the histogram is bunched up in one area. I also note that all the dark areas have come up well with information, similar to in the original image. It is a good demonstration of what can be 'brought back' from a seemingly very dark image! I also display the curve adjustment that I applied to the above image.

Resultant image

Curves applied to simulate what camera does automatically

This exercise shows the importance of correct exposure in-camera. Avoiding underexposure is especially important because any noise in the dark areas will be exacerbated by the automatic in-camera processing. It also shows me what can be brought back from an image which may be useful for some images that are accidentally poorly exposed.

No comments:

Post a Comment