Talk:Image gradient

Confusing
I found the definition of the gradient to need further clarification:

The gradient of an image is given by the formula :


 * $$\nabla f=\frac{\partial f}{\partial x}\hat x + \frac{\partial f}{\partial y}\hat y$$,

where :


 * $$\textstyle\frac{\partial f}{\partial x}$$ is the gradient in the x direction


 * $$\textstyle\frac{\partial f}{\partial y}$$ is the gradient in the y direction.

For instance:

What is $$\hat x$$?

What is $$\hat y$$?

Are these unit vectors?

And, should there be a link to computing the discrete gradient for the definition of the gradient for the x and y directions?

I was honestly confused here.

FAdmMatt (talk) 07:46, 28 January 2014 (UTC)

From white to black
This article contravenes the convention that "dark" image regions represent "low" values and "bright" image regions represent "high" values. Should simply reverse the sense of these words and the accompanying graphic. —Preceding unsigned comment added by Kennyhunt (talk • contribs) 16:24, August 17, 2006


 * This has recently confused a poster to the sci.image.processing newsgroup, so this certainly needs fixing. I suggest somebody copy the images used and take their negative. Then it will make sense. HairyWombat (talk) 14:21, 1 July 2009 (UTC)

merge from color progression
These two articles give different names to the same thing. As gradient is the more common term, they should be merged here, and color progression change to a redirect. --jacobolus (t) 21:56, 20 September 2007 (UTC)


 * Changed #REDIRECT at Color progression from this article to the article Color gradient. This is more appropriate, as this article isn't really anything to do with colour. HairyWombat (talk) 01:53, 2 July 2009 (UTC)

inaccurate ??
"The Sobel operator represents a rather inaccurate approximation of the image gradient"

The image gradient is the gradient of an assumed (not uniquely defined) function. What does it mean to be inaccurate? Surely we can construct a continuous function that takes on the discrete pixel values and matches the sobel derivative, then the sobel value would be exact and anything else would be inaccurate.

That statement needs to be cited and explained or removed. Likewise "It turns out that the derivatives at any particular point are functions of the intensity values at virtually all image points" is nonsense without some further assumption about the nature of the assumed continuous function.

99.56.163.151 (talk) 23:23, 19 March 2012 (UTC)
 * I tend to agree with both of these criticisms - from what I have read, Sobel is not an accurate way to detect edges... but edge detection is not what this article is about. And as you mentioned, the gradient is not globally dependent in the general case. I will remove the second sentence and rewrite the first. Smitty121981 (talk) 15:21, 8 August 2014 (UTC)