Edge Detector Info?

(note im not talking about silloettes)

u can do a dotproduct( view_dir, normal ) if its close to 0.0 its gonna be on the edge

or u can ( i was testing AA methods with this method ) compare the pixels depth value (from the depthbuffer) with its neightbours depth value if the differnce is over a certain value thens its at an edge

u could also do a sharpening filter

Ok, edge detecting what? For colour images, there's the Robust Colour Morphological Gradient - which produces decent results and a relatively low computational cost. There's also the compass method which does produce better results than the RCMG - however it's far more computationally expensive. If you want to know more about the RCMG and the CM operations required then let me know.

For greyscale, there's a few decent ones. Mainly, these are the Laplacian of Gaussian (LoG) and Canny edge detectors.

python_regious's reply is pretty complete, but for completeness: The Sobel operator is not entirely useless on noisy images. If you 'close' the image by eroding it then dilating it, noise will be greatly reduced, without destroying the significant details. The technique can be applied several times if necessary (at the cost of image quality) before applying a Sobel filter. However, in many cases, there are better alternatives.

As for blurry images: they are, by definition, relatively noise-free, so you have a broad range of filters at your disposal, Sobel included.

Regards
Admiral