## -*- mode: tcl ; fill-column: 90 -*-
# # ## ### ##### ######## ############# #####################
## Getter -- Threshold generation - Global
#
# https://craftofcoding.wordpress.com/2021/10/27/thresholding-algorithms-bernsen-local/
# https://craftofcoding.wordpress.com/2021/09/30/thresholding-algorithms-niblack-local/
# https://craftofcoding.wordpress.com/2021/10/06/thresholding-algorithms-sauvola-local/
# https://craftofcoding.wordpress.com/2021/09/28/thresholding-algorithms-phansalkar-local/
# # ## ### ##### ######## ############# #####################
## Mean
def ref-bernsen \[Bernsen\](https://craftofcoding.wordpress.com/2021/10/27/thresholding-algorithms-bernsen-local)
def ref-sauvola \[Sauvola\](https://craftofcoding.wordpress.com/2021/10/06/thresholding-algorithms-sauvola-local)
def ref-niblack \[Niblack\](https://craftofcoding.wordpress.com/2021/09/30/thresholding-algorithms-niblack-local)
def ref-phansalkar \[Phansalkar\](https://craftofcoding.wordpress.com/2021/09/28/thresholding-algorithms-phansalkar-local)
def ref-otsu \[Otsu\](https://en.wikipedia.org/wiki/Otsu%27s_method)
operator image::threshold::global::mean {
section accessor threshold generate
example {
scancrop
@1 | -text
}
example {
butterfly
@1 | -text
}
note Returns a global threshold for the input, as the image mean.
note The operator "<!xref: aktive image mask per global mean>" \
uses this to generate a mask of the input.
note There are better methods. Extensions to the simple mean, in order \
of creation (and complexity), are @@ref-sauvola@@, @@ref-niblack@@, and \
@@ref-phansalkar@@. Each of these modifies the plain mean with a bias \
based on a mix of mean, standard deviation, and parameters.
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = meanI
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# T = meanI - src
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
return [aktive op image mean $src]
}
}
# # ## ### ##### ######## ############# #####################
## Bernsen
operator image::threshold::global::bernsen {
section accessor threshold generate
example {
scancrop
@1 | -text
}
example {
butterfly
@1 | -text
}
note Returns a global threshold for the input, \
according to @@ref-bernsen@@'s method.
note The operator "<!xref: aktive image mask per global bernsen>" \
uses this to generate a mask of the input.
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = (minI + maxI) / 2
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# maxI
# / \.
# T = (*) - (+) >- src
# \ \. /
# 0.5 maxI
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
set min [aktive op image min $src]
set max [aktive op image max $src]
return [expr {0.5 * ($min + $max)}]
}
}
# # ## ### ##### ######## ############# #####################
## Niblack
operator image::threshold::global::niblack {
section accessor threshold generate
example {
scancrop
@1 | -text
}
example {
butterfly
@1 | -text
}
note Returns a global threshold for the input, \
according to @@ref-niblack@@'s method.
note The operator "<!xref: aktive image mask per global niblack>" \
uses this to generate a mask of the input.
double? -0.2 k niblack parameter
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = meanI + (k * stdI)
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# meanI ---\.
# / \.
# t = (+) /- src
# \. /
# (*) - stdI
# \.
# k
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
set mean [aktive op image mean $src]
set std [aktive op image stddev $src]
return [expr {$mean + $k * $std}]
}
}
# # ## ### ##### ######## ############# #####################
## Sauvola
operator image::threshold::global::sauvola {
section accessor threshold generate
example {
scancrop
@1 | -text
}
example {
butterfly
@1 | -text
}
note Returns a global threshold for the input, \
according to @@ref-sauvola@@'s method.
note The operator "<!xref: aktive image mask per global sauvola>" \
uses this to generate a mask of the input.
double? 0.5 k sauvola parameter
double? 128 R sauvola parameter
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = meanI * (1 + k * ((stdI/R) - 1))
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# meanI -------------------\.
# / /- src
# t = (*) (*) - (-) - (/) - stdI
# \. / \ \ \.
# (+) k 1 R
# \.
# 1
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
set mean [aktive op image mean $src]
set std [aktive op image stddev $src]
return [expr {$mean * (1. + $k * ($std / double($R) - 1.))}]
}
}
# # ## ### ##### ######## ############# #####################
## Phansalkar
operator image::threshold::global::phansalkar {
section accessor threshold generate
example {
scancrop
@1 | -text
}
example {
butterfly
@1 | -text
}
note Returns a global threshold for the input, \
according to @@ref-phansalkar@@'s method.
note The operator "<!xref: aktive image mask per global phansalkar>" \
uses this to generate a mask of the input.
double? 0.25 k phansalkar parameter
double? 0.5 R phansalkar parameter
double? 3 p phansalkar parameter
double? 10 q phansalkar parameter
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = meanN * (p * exp(-q * meanN) + 1 + k * ((stdN / R) - 1))
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# /------------------- meanN
# / / \.
# / (exp) - (*) \.
# / / \ -q \- src
# / (*) - p /
# t = (*) / /
# \- (+) - (*) - (-) - (/) - stdN
# \ \ \ \.
# 1 k -1 R
#
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
set mean [aktive op image mean $src]
set std [aktive op image stddev $src]
return [expr {$mean * ($p * exp(- $q * $mean) + 1. + $k * ($std / double($R) - 1.))}]
}
}
# # ## ### ##### ######## ############# #####################
## Otsu
operator image::threshold::global::otsu {
section accessor threshold generate
example {
scancrop
@1 | -text
}
note Returns a global threshold for the input, \
according to @@ref-otsu@@'s method.
note The operator "<!xref: aktive image mask per global otsu>" \
uses this to generate a mask of the input.
int? 256 bins \
The number of bins used by the internal histogram. \
The pixel values are quantized to fit. \
Only values in the range of `\[0..1\]` are considered valid. \
Values outside of that range are placed into the smallest/largest \
bins, respectively. \
\
The default quantizes the image values to 8-bit.
input
strict single \
The computed pixels are not materialized. \
They are immediately reduced to the threshold.
body {
# t = scaled (extract (row otsu (image histogram I)))
# the threshold is scaled by the number of bins, to be in [0..1]
set e [aktive op image histogram $src bins $bins]
set e [aktive op row otsu $e]
return [expr {[lindex [aktive query values $e] 0] / double($bins)}]
}
}
##
# # ## ### ##### ######## ############# #####################
::return