Convolution Kernel: Difference between revisions

(3 intermediate revisions by the same user not shown)

Line 1:

[[Category:Techniques]]

~~hope you know linear algebra!~~

A '''Convolution Kernel''' is when you use specific matrix transformations (known as convolutions) into a image's kernel to create a wide variety of effects.

A '''Convolution Kernel''' is when you use specific matrix transformations (known as convolutions) into a image's kernel to create a wide variety of effects. Working with these kernels can create a wide variety of effects, from blurs to sharpens and more.

While many effects in modern editors stem from a convolution kernel they usually limit editing to just that one style of effect; working with convolution kernels ''directly'' is rare as it often requires a linear algebra background and the manual editing of many matrix points.

=Effect Types=

==Emboss==

[[File:Yasmine emboss.png|200px]]

One of the most well known and instantly recognizable convolution patterns, instantly recognizable from its gray coloring and black outlines. Available as a preset in almost every major editor, including [[Premiere]], [[Vegas]] and [[Kdenlive]].

==Edge Detection==

[[File:Yasmine edgedetect.png|200px]]

Another well known filter that emphasizes the "edges" of an image and removes almost everything else.

==Sharpen==

Similar to edge detection, sharpen will emphasize edges - without removing other details.

==Gaussian Blur==

Line 13:

Line 26:

==Unsharp Mask==

~~Very~~ popular in ~~recent years~~.

An '''unsharp mask''' is a type of convolution kernel where the image is masked with a copy that is blurred (or "un-sharpened"). This can make a clearer image that may have less detail.

Unsharp masks have been a very popular choice in [[tennis]] in the late 2010's, popularized by players like {{mafar}}.

@@ Line 1: / Line 1: @@
 [[Category:Techniques]]
-hope you know linear algebra!
+{{Stub}}
-A '''Convolution Kernel''' is when you use specific matrix transformations (known as convolutions) into a image's kernel to create a wide variety of effects.
+A '''Convolution Kernel''' is when you use specific matrix transformations (known as convolutions) into a image's kernel to create a wide variety of effects. Working with these kernels can create a wide variety of effects, from blurs to sharpens and more.
+While many effects in modern editors stem from a convolution kernel they usually limit editing to just that one style of effect; working with convolution kernels ''directly'' is rare as it often requires a linear algebra background and the manual editing of many matrix points.
+=Effect Types=
+==Emboss==
+[[File:Yasmine emboss.png|200px]]
+One of the most well known and instantly recognizable convolution patterns, instantly recognizable from its gray coloring and black outlines. Available as a preset in almost every major editor, including [[Premiere]], [[Vegas]] and [[Kdenlive]].
 ==Edge Detection==
+[[File:Yasmine edgedetect.png|200px]]
+Another well known filter that emphasizes the "edges" of an image and removes almost everything else.
 ==Sharpen==
+Similar to edge detection, sharpen will emphasize edges - without removing other details.
 ==Gaussian Blur==
@@ Line 13: / Line 26: @@
 ==Unsharp Mask==
-Very popular in recent years.
+An '''unsharp mask''' is a type of convolution kernel where the image is masked with a copy that is blurred (or "un-sharpened"). This can make a clearer image that may have less detail.
+Unsharp masks have been a very popular choice in [[tennis]] in the late 2010's, popularized by players like {{mafar}}.