Blur
(→Developer notes) |
|||
Line 41: | Line 41: | ||
==== Developer notes ==== | ==== Developer notes ==== | ||
'''Blur''' uses the kernel [(1−1/2^amount)/2, 1/2^amount, (1−1/2^amount)/2]. The largest allowable argument for Blur is log2(3) (which is about 1.58), which corresponds to a (1/3,1/3,1/3) kernel. A value of 1.0 gets you a (1/4,1/2,1/4) kernel for example. | '''Blur''' uses the kernel [(1−1/2^amount)/2, 1/2^amount, (1−1/2^amount)/2]. The largest allowable argument for Blur is log2(3) (which is about 1.58), which corresponds to a (1/3,1/3,1/3) kernel. A value of 1.0 gets you a (1/4,1/2,1/4) kernel for example. | ||
− | Likewise Blur(1.0).Blur(1.0) is a convolution of the kernel (1/4,1/2,1/4) with itself, | + | Likewise Blur(1.0).Blur(1.0) is a convolution of the kernel (1/4,1/2,1/4) with itself, being a (1/4,1/2,1/4)*(1/4,1/2,1/4) = (1/16,4/16,6/16,4/16,1/16) kernel. It can be read of [https://en.wikipedia.org/wiki/Pascal's_triangle Pascal's triangle]. |
− | + | ||
==== Changes ==== | ==== Changes ==== |
Revision as of 19:44, 16 March 2016
Blur(clip clip, float amount, bool MMX)
Blur(clip, float amountH, float amountV, bool MMX)
- A 3×3 kernel blurring filter.
Sharpen(clip clip, float amount, bool MMX)
Sharpen(clip, float amountH, float amountV, bool MMX)
- A 3×3 kernel sharpening filter; the inverse of Blur.
float amount = (required)
- The allowable range for Blur is from -1.0 to +1.58
- The allowable range for Sharpen is from -1.58 to +1.0
- Negative Blur actually sharpens the image; in fact Sharpen(n) is just an alias for Blur(-n).
float amountH = (required)
float amountV = (amountH)
- You can use 2 arguments to set independent vertical and horizontal blurring or sharpening: for example,
Blur(0,1)
- will blur vertical only, perhaps to blend interlaced lines together.
bool MMX = true
- This option should always be true.
Notes
If you need a larger radius Gaussian blur, try chaining several Blurs together:
Blur(1.0).Blur(1.0).Blur(1.0)
Chaining calls to Sharpen is not a good idea, as the image quickly deteriorates.
Developer notes
Blur uses the kernel [(1−1/2^amount)/2, 1/2^amount, (1−1/2^amount)/2]. The largest allowable argument for Blur is log2(3) (which is about 1.58), which corresponds to a (1/3,1/3,1/3) kernel. A value of 1.0 gets you a (1/4,1/2,1/4) kernel for example. Likewise Blur(1.0).Blur(1.0) is a convolution of the kernel (1/4,1/2,1/4) with itself, being a (1/4,1/2,1/4)*(1/4,1/2,1/4) = (1/16,4/16,6/16,4/16,1/16) kernel. It can be read of Pascal's triangle.
Changes
v2.58 | MMX routines fixed (have full 8 bit precision now); mmx=true by default |
v2.57 | added MMX option |