42 lines
1.1 KiB
Plaintext
42 lines
1.1 KiB
Plaintext
|
= Gradient Vectors =
|
||
|
|
|
||
|
|
A gradiant vector is an array with one cell corresponding to each pixel in an
|
||
|
|
image, with each cell containing a [[vector]] of the image's gradiant at that
|
||
|
|
point.
|
||
|
|
|
||
|
|
A Vector in this case has both direction and maginiute.
|
||
|
|
|
||
|
|
A gradient in this case is the change in values along both the x and y around
|
||
|
|
some pixel.
|
||
|
|
|
||
|
|
An interesting and very important property of gradient vectors is their
|
||
|
|
relationship to edges in an image. An edge is any sudden change in an image,
|
||
|
|
including lines, boundaries between two objects, etc. A gradient vector will
|
||
|
|
always be a normal to the edge in an image.
|
||
|
|
|
||
|
|
== Example Algorithm ==
|
||
|
|
|
||
|
|
{{{
|
||
|
|
[X] [7] [X]
|
||
|
|
[2] [C] [8]
|
||
|
|
[X] [3] [X]
|
||
|
|
}}}
|
||
|
|
|
||
|
|
Given the above pixel C, we only care about the values with numbers. We can
|
||
|
|
find the gradient in the X and Y respectively using
|
||
|
|
|
||
|
|
{{{
|
||
|
|
Delta_x = 8 - 2 = 6
|
||
|
|
Delta_y = 7 - 3 = 4
|
||
|
|
}}}
|
||
|
|
|
||
|
|
Therefore our vector is the matrix consisting of 6 over 4. To find the
|
||
|
|
direction of this vector we can take
|
||
|
|
|
||
|
|
{{{
|
||
|
|
arctan(6 / 4) = 56.309 degrees
|
||
|
|
}}}
|
||
|
|
|
||
|
|
This of course was on a single pixel value but the same concept applies for [[RGB]]
|
||
|
|
or [[CMYK]] colorspaces.
|