Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
I
imagej-elphel
Project
Project
Details
Activity
Releases
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
3
Issues
3
List
Board
Labels
Milestones
Wiki
Wiki
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Commits
Issue Boards
Open sidebar
Elphel
imagej-elphel
Commits
ab8732b9
Commit
ab8732b9
authored
Jan 03, 2023
by
Andrey Filippov
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
some improvement in trimming
parent
0ee312b7
Changes
2
Show whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
624 additions
and
42 deletions
+624
-42
TexturedModel.java
...n/java/com/elphel/imagej/tileprocessor/TexturedModel.java
+291
-42
TileNeibs.java
src/main/java/com/elphel/imagej/tileprocessor/TileNeibs.java
+333
-0
No files found.
src/main/java/com/elphel/imagej/tileprocessor/TexturedModel.java
View file @
ab8732b9
...
...
@@ -35,6 +35,7 @@ import org.json.JSONException;
import
com.elphel.imagej.cameras.CLTParameters
;
import
com.elphel.imagej.cameras.ColorProcParameters
;
import
com.elphel.imagej.cameras.EyesisCorrectionParameters
;
import
com.elphel.imagej.common.DoubleGaussianBlur
;
import
com.elphel.imagej.common.ShowDoubleFloatArrays
;
import
com.elphel.imagej.correction.EyesisCorrections
;
import
com.elphel.imagej.x3d.export.GlTfExport
;
...
...
@@ -3235,15 +3236,14 @@ public class TexturedModel {
final
boolean
[][]
seed_pix
,
// FG edge, just outside of trim_pix. Will be modified
final
double
[][]
vars_same
,
final
double
[][]
vars_inter
,
final
double
seed_
inter
,
// minimal value of vars_inte
r
final
double
seed_
same_fz
,
// add to var_same in denominato
r
final
double
seed_fom
,
// minimal value of vars_inter/sqrt(vars_same)
final
double
seed_inter
,
// = 150;
final
int
width
)
{
final
int
num_slices
=
trim_pix
.
length
;
final
int
num_pix
=
trim_pix
[
0
].
length
;
final
int
height
=
num_pix
/
width
;
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
final
TileNeibs
pn
=
new
TileNeibs
(
width
,
height
);
for
(
int
nslice
=
0
;
nslice
<
num_slices
;
nslice
++)
{
final
int
fnslice
=
nslice
;
ai
.
set
(
0
);
...
...
@@ -3251,8 +3251,10 @@ public class TexturedModel {
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
num_pix
;
pix
=
ai
.
getAndIncrement
())
if
(
trim_pix
[
fnslice
][
pix
])
{
if
(
vars_inter
[
fnslice
][
pix
]
>=
seed_inter
)
{
double
fom
=
vars_inter
[
fnslice
][
pix
]/
Math
.
sqrt
(
vars_same
[
fnslice
][
pix
]);
if
(
vars_inter
[
fnslice
][
pix
]
>
seed_inter
)
{
// works for high-contrast over sky
seed_pix
[
fnslice
][
pix
]
=
true
;
}
else
{
// for foreground with lower contrast
double
fom
=
vars_inter
[
fnslice
][
pix
]/(
vars_same
[
fnslice
][
pix
]
+
seed_same_fz
);
if
(
fom
>=
seed_fom
)
{
seed_pix
[
fnslice
][
pix
]
=
true
;
}
...
...
@@ -3266,13 +3268,90 @@ public class TexturedModel {
return
seed_pix
;
}
public
static
boolean
[][]
getTrimAlpha
(
public
static
double
[][]
getTrimFom
(
final
boolean
[][]
trim_pix
,
// pixels that may be trimmed
final
boolean
[][]
seed_pix
,
// FG edge (just outside of trim_pix) and seeds from vars_inter mismatch
final
double
[][]
vars_same
,
final
double
[][]
vars_inter
,
final
double
thr_same
,
// minimal value of vars_same to block propagation
final
double
thr_ratio
,
// minimal value of vars_same/vars_inter to block propagation
final
double
trim_inter_fz
,
// minimal value of vars_same to block propagation
final
double
trim_fom_threshold
,
// = 120.0; // count only pixels with VAR_SAME > this value
final
double
trim_fom_boost
,
// = 5; // boost high-varinace values that exceed threshold
final
double
trim_fom_blur
,
final
int
width
,
final
double
[][][]
fom_dbg
)
{
final
int
num_slices
=
trim_pix
.
length
;
final
int
num_pix
=
trim_pix
[
0
].
length
;
final
int
height
=
num_pix
/
width
;
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
final
double
[][]
trim_foms
=
new
double
[
num_slices
][
num_pix
];
final
double
[]
fom_blurred
=
(
trim_fom_blur
>
0
)
?
new
double
[
num_pix
]
:
null
;
final
double
boost_scale
=
trim_fom_boost
/
trim_fom_threshold
;
final
double
boost_subtract
=
trim_fom_boost
-
1.0
;
if
(
fom_dbg
!=
null
)
{
for
(
int
i
=
0
;
i
<
fom_dbg
.
length
;
i
++)
{
fom_dbg
[
i
]
=
new
double
[
num_slices
][];
}
}
for
(
int
nslice
=
0
;
nslice
<
num_slices
;
nslice
++)
{
final
int
fnslice
=
nslice
;
Arrays
.
fill
(
trim_foms
[
fnslice
],
Double
.
NaN
);
if
(
trim_fom_blur
>
0
)
{
Arrays
.
fill
(
fom_blurred
,
1.0
);
// trim_fom_threshold);
}
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
num_pix
;
pix
=
ai
.
getAndIncrement
())
if
(!
Double
.
isNaN
(
vars_same
[
fnslice
][
pix
])){
if
(
trim_pix
[
fnslice
][
pix
])
{
// should be NaN outside of trim_pix
trim_foms
[
fnslice
][
pix
]
=
vars_same
[
fnslice
][
pix
]/(
vars_inter
[
fnslice
][
pix
]
+
trim_inter_fz
);
}
if
((
trim_fom_blur
>
0
)
&&
!
Double
.
isNaN
(
trim_foms
[
fnslice
][
pix
]))
{
fom_blurred
[
pix
]
=
Math
.
max
(
vars_same
[
fnslice
][
pix
],
trim_fom_threshold
)
*
boost_scale
-
boost_subtract
;
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
if
(
fom_dbg
!=
null
)
{
fom_dbg
[
0
][
nslice
]
=
trim_foms
[
fnslice
].
clone
();
fom_dbg
[
2
][
nslice
]
=
fom_blurred
.
clone
();
}
if
(
trim_fom_blur
>
0
)
{
(
new
DoubleGaussianBlur
()).
blurDouble
(
fom_blurred
,
width
,
height
,
trim_fom_blur
,
trim_fom_blur
,
0.01
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
num_pix
;
pix
=
ai
.
getAndIncrement
())
if
(!
Double
.
isNaN
(
trim_foms
[
fnslice
][
pix
])){
trim_foms
[
fnslice
][
pix
]
/=
fom_blurred
[
pix
];
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
if
(
fom_dbg
!=
null
)
{
fom_dbg
[
3
][
nslice
]
=
fom_blurred
.
clone
();
fom_dbg
[
1
][
nslice
]
=
trim_foms
[
fnslice
].
clone
();
}
}
}
return
trim_foms
;
}
public
static
void
getTrimAlpha
(
final
double
[][]
fom_pix
,
// should be NaN outside of trim_pix
final
boolean
[][]
trim_pix
,
// pixels that may be trimmed
final
boolean
[][]
seed_pix
,
// FG edge (just outside of trim_pix) and seeds from vars_inter mismatch
final
double
trim_fom
,
// minimal value of vars_same/vars_inter to block propagation
final
int
trim_grow
,
// 3*transform_size?
final
int
width
)
{
final
int
num_slices
=
trim_pix
.
length
;
...
...
@@ -3282,6 +3361,7 @@ public class TexturedModel {
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
final
TileNeibs
pn
=
new
TileNeibs
(
width
,
height
);
final
double
[][]
foms
=
new
double
[
2
][
num_pix
];
for
(
int
nslice
=
0
;
nslice
<
num_slices
;
nslice
++)
{
final
int
fnslice
=
nslice
;
Arrays
.
fill
(
prohibit
,
true
);
...
...
@@ -3290,14 +3370,11 @@ public class TexturedModel {
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
num_pix
;
pix
=
ai
.
getAndIncrement
())
{
if
(
trim_pix
[
fnslice
][
pix
])
{
if
((
vars_same
[
fnslice
][
pix
]
<
thr_same
)
||
(
vars_same
[
fnslice
][
pix
]/
vars_inter
[
fnslice
][
pix
]
<
thr_ratio
)
)
{
// NaN should keep prohibit (fom_pix should be NaN outside of trim_pix
if
(((
fom_pix
[
fnslice
][
pix
]
<=
trim_fom
)
&&
trim_pix
[
fnslice
][
pix
])
||
seed_pix
[
fnslice
][
pix
]
)
{
prohibit
[
pix
]
=
false
;
}
}
else
if
(
seed_pix
[
fnslice
][
pix
])
{
prohibit
[
pix
]
=
false
;
// enable FG edge that is the only outside trim_pix
}
}
}
};
...
...
@@ -3315,10 +3392,59 @@ public class TexturedModel {
seed_pix
[
fnslice
],
seed_pix
[
fnslice
]);
}
return
seed_pix
;
// extends outside selected tiles, but that's OK
return
;
// seed_pix; // extends outside selected tiles, but that's OK
}
public
static
void
expandTrimAlpha
(
final
boolean
[][]
trim_pix
,
// pixels that may be trimmed
final
boolean
[][]
alpha_pix
,
//
final
double
[][]
value
,
// will grow only in increasing
final
double
min_incr
,
final
boolean
dual_pass
,
final
int
trim_grow
,
// 3*transform_size?
final
int
width
)
{
final
int
num_slices
=
trim_pix
.
length
;
final
int
num_pix
=
trim_pix
[
0
].
length
;
final
int
height
=
num_pix
/
width
;
final
boolean
[]
removed
=
new
boolean
[
num_pix
];
final
boolean
[]
prohibit
=
new
boolean
[
num_pix
];
final
TileNeibs
pn
=
new
TileNeibs
(
width
,
height
);
for
(
int
nslice
=
0
;
nslice
<
num_slices
;
nslice
++)
{
final
int
fnslice
=
nslice
;
TileNeibs
.
invertSelection
(
alpha_pix
[
fnslice
],
removed
);
//
TileNeibs
.
invertSelection
(
trim_pix
[
fnslice
],
prohibit
);
// prohibit all that is not trim
pn
.
growSelectionGradient
(
//
trim_grow
,
// int grow, // grow tile selection by 1 over non-background tiles 1: 4 directions, 2 - 8 directions, 3 - 8 by 1, 4 by 1 more
removed
,
// final boolean [] tiles,
prohibit
,
// final boolean [] prohibit,
value
[
fnslice
],
// final double [] value,
min_incr
,
// final double min_incr,
true
,
// final boolean keep_top, // do not grow over local max
false
);
// final boolean keep_thin_top)// do not grow over local max only if next after that is already selected
if
(
dual_pass
)
{
pn
.
growSelectionGradient
(
//
2
,
// trim_grow, // int grow, // grow tile selection by 1 over non-background tiles 1: 4 directions, 2 - 8 directions, 3 - 8 by 1, 4 by 1 more
removed
,
// final boolean [] tiles,
prohibit
,
// final boolean [] prohibit,
value
[
fnslice
],
// final double [] value,
min_incr
,
// final double min_incr,
false
,
// final boolean keep_top, // do not grow over local max
true
);
// final boolean keep_thin_top)// do not grow over local max only if next after that is already selected
}
TileNeibs
.
invertSelection
(
removed
,
alpha_pix
[
fnslice
]);
//
}
return
;
}
public
static
void
filterAlpha
(
final
boolean
[][]
alpha_pix
,
// per-pixel alpha
final
boolean
[][]
trim_pix
,
// pixels that may be trimmed
...
...
@@ -3684,7 +3810,6 @@ public class TexturedModel {
final
double
[][]
combo_texture
,
final
double
var_radius
,
final
int
width
)
{
// last slice - ratio
final
int
num_slices
=
sensor_texture
.
length
;
final
int
num_sensors
=
sensor_texture
[
0
].
length
;
final
int
img_size
=
sensor_texture
[
0
][
0
].
length
;
...
...
@@ -3696,7 +3821,7 @@ public class TexturedModel {
var_weights
[
i
][
j
]
=
Math
.
cos
(
0.5
*
Math
.
PI
*
i
/
var_radius
)
*
Math
.
cos
(
0.5
*
Math
.
PI
*
j
/
var_radius
);
}
}
final
double
[][][]
vars
=
new
double
[
2
][
num_slices
][
img_size
];
final
double
[][][]
vars
=
new
double
[
5
][
num_slices
][
img_size
];
// same, inter, d/dx, d/dy,sqrt ((d/dx)^2+(d/dy)^2)
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
final
TileNeibs
pn
=
new
TileNeibs
(
width
,
height
);
...
...
@@ -3717,24 +3842,32 @@ public class TexturedModel {
double
var_same
=
Double
.
NaN
;
double
var_inter
=
Double
.
NaN
;
// calculate weighted variance
double
sw
=
0.0
,
swd
=
0.0
,
swd2
=
0.0
;
double
sw
=
0.0
,
swd
=
0.0
,
swd2
=
0.0
,
swxd
=
0.0
,
swyd
=
0.0
;
for
(
int
dvy
=
-
ivar_radius
;
dvy
<=
ivar_radius
;
dvy
++)
{
for
(
int
dvx
=
-
ivar_radius
;
dvx
<=
ivar_radius
;
dvx
++)
{
int
indx
=
pn
.
getIndex
(
x0
+
dvx
,
y0
+
dvy
);
if
((
indx
>=
0
)
&&
!
Double
.
isNaN
(
combo_texture
[
fnslice
][
indx
]))
{
double
w
=
var_weights
[
Math
.
abs
(
dvy
)][
Math
.
abs
(
dvx
)];
// 1.0;
double
d
=
combo_texture
[
fnslice
][
indx
];
double
wd
=
w
*
d
;
sw
+=
w
;
swd
+=
w
*
d
;
swd2
+=
w
*
d
*
d
;
swd
+=
wd
;
swd2
+=
wd
*
d
;
swxd
+=
dvx
*
wd
;
swyd
+=
dvy
*
wd
;
}
}
}
if
(
sw
>
0.0
)
{
// always
double
avg
=
swd
/
sw
;
double
avg2
=
swd2
/
sw
;
swxd
/=
sw
;
swyd
/=
sw
;
var_same
=
Math
.
sqrt
(
avg2
-
avg
*
avg
);
vars
[
0
][
fnslice
][
cindx
]
=
var_same
;
vars
[
2
][
fnslice
][
cindx
]
=
swxd
;
vars
[
3
][
fnslice
][
cindx
]
=
swyd
;
vars
[
4
][
fnslice
][
cindx
]
=
Math
.
sqrt
(
swxd
*
swxd
+
swyd
*
swyd
);
}
// calculate inter-sensor variance (add local normalization?)
sw
=
0.0
;
swd
=
0.0
;
swd2
=
0.0
;
...
...
@@ -4898,7 +5031,7 @@ public class TexturedModel {
final
double
max_disparity_lim
,
// = 100.0; // do not allow stray disparities above this
final
double
min_trim_disparity
,
// = 2.0; // do not try to trim texture outlines with lower disparities
final
String
dbg_prefix
)
{
final
double
var_radius
=
3.5
;
// for variance filter of the combo disparity
final
double
var_radius
=
1.5
;
//
3.5; // for variance filter of the combo disparity
final
double
dir_radius
=
1.5
;
// averaging inter-sensor variance to view behind obstacles
final
double
try_dir_var
=
20.0
;
// try directional if the intersensor variance exceeds this value
final
int
dir_num_start
=
7
;
// start with this number of consecutive sensors
...
...
@@ -4933,13 +5066,22 @@ public class TexturedModel {
// New processing
final
double
seed_inter
=
50.0
;
final
double
seed_fom
=
15.0
;
final
double
thr_same
=
16
;
// 20; // minimal value of vars_same to block propagation
final
double
thr_ratio
=
2.5
;
// 3.0; // minimal value of vars_same/vars_inter to block propagation
final
double
seed_inter
=
150
;
// 120; // 150;
final
double
seed_same_fz
=
6.5
;
// 13; // seed_inter = 50.0;
final
double
seed_fom
=
2.0
;
// 1.9; // 1.2;
final
double
trim_inter_fz
=
5.0
;
// 13.0;
final
double
trim_fom
=
0.5
;
// 0.8; // 1.3; // 1.8; // 1.2; // 0.8; // 0.4; // 0.7;
// scale down fom for pixels near high-variance VAR_SAME
final
double
trim_fom_threshold
=
120.0
;
// count only pixels with VAR_SAME > this value
final
double
trim_fom_boost
=
5
;
// boost high-varinace values that exceed threshold
final
double
trim_fom_blur
=
10.0
;
// divide trim_fom array by blurred version to reduce over sky sharp edge
final
double
min_incr
=
100
;
// temporary disable // 5; // 20.0; // 0.5; // only for sky?
// final double thr_same = 16; // 20; // minimal value of vars_same to block propagation
// final double thr_ratio = 2.5; // 3.0; // minimal value of vars_same/vars_inter to block propagation
final
int
trim_grow_pix
=
transform_size
*
3
;
// 3*transform_size?
final
int
min_neibs_alpha
=
1
;
// minimal neighbors to keep alpha
final
int
grow_alpha
=
2
;
// grow alpha selection
final
int
grow_alpha
=
0
;
//
2; // grow alpha selection
final
double
alphaOverlapTolerance
=
0.0
;
// exact match only
final
int
reduce_has_bg_grow
=
2
;
// 0 - exactly half tile (between strong and weak)
// final int strong_bg_overlap = 1;
...
...
@@ -5034,8 +5176,9 @@ public class TexturedModel {
unbound_alpha
,
// final boolean [][] seed_pix, // FG edge, just outside of trim_pix. Will be modified
vars
[
0
],
// final double [][] vars_same,
vars
[
1
],
// final double [][] vars_inter,
seed_
inter
,
// final double seed_inter, // minimal value of vars_inte
r
seed_
same_fz
,
// final double seed_same_fz, // add to var_same in denominato
r
seed_fom
,
// final double seed_fom, // minimal value of vars_inter/sqrt(vars_same)
seed_inter
,
// final double seed_inter, // = 150;
width
);
// final int width)
// copy unbound_alpha here for debug
// boolean [][] unbound_alpha =
...
...
@@ -5047,16 +5190,42 @@ public class TexturedModel {
trim_seeds
[
i
]
=
unbound_alpha
[
i
].
clone
();
}
}
getTrimAlpha
(
final
double
[][][]
fom_dbg
=
(
dbg_prefix
!=
null
)?
new
double
[
4
][][]
:
null
;
double
[][]
trim_fom_pix
=
getTrimFom
(
trim_pixels
,
// final boolean [][] trim_pix, // pixels that may be trimmed
unbound_alpha
,
// final boolean [][] seed_pix, // FG edge (just outside of trim_pix) and seeds from vars_inter mismatch
vars
[
0
],
// final double [][] vars_same,
vars
[
1
],
// final double [][] vars_inter,
thr_same
,
// final double thr_same, // minimal value of vars_same to block propagation
thr_ratio
,
// final double thr_ratio, // minimal value of vars_same/vars_inter to block propagation
trim_inter_fz
,
// final double trim_inter_fz, // minimal value of vars_same to block propagation
trim_fom_threshold
,
//final double trim_fom_threshold, // = 120.0; // count only pixels with VAR_SAME > this value
trim_fom_boost
,
//final double trim_fom_boost, // = 5; // boost high-varinace values that exceed threshold
trim_fom_blur
,
// final double trim_fom_blur,
width
,
// final int width)
fom_dbg
);
// final double [][][] fom_dbg)
getTrimAlpha
(
trim_fom_pix
,
// final double [][] fom_pix, // should be NaN outside of trim_pix
trim_pixels
,
// final boolean [][] trim_pix, // pixels that may be trimmed
unbound_alpha
,
// final boolean [][] seed_pix, // FG edge (just outside of trim_pix) and seeds from vars_inter mismatch
trim_fom
,
// final double trim_fom, // minimal value of vars_same/vars_inter to block propagation
trim_grow_pix
,
// final int trim_grow, // 3*transform_size?
width
);
// final int width)
final
boolean
[][]
first_trimmed_alpha
=
(
dbg_prefix
!=
null
)?
new
boolean
[
num_slices
][]
:
null
;
if
(
dbg_prefix
!=
null
)
{
for
(
int
i
=
0
;
i
<
num_slices
;
i
++)
{
first_trimmed_alpha
[
i
]
=
unbound_alpha
[
i
].
clone
();
}
}
final
boolean
dual_pass
=
false
;
// true;
expandTrimAlpha
(
trim_pixels
,
// final boolean [][] trim_pix, // pixels that may be trimmed
unbound_alpha
,
// final boolean [][] alpha_pix, //
vars
[
0
],
// final double [][] value, // will grow only in increasing
min_incr
,
// final double min_incr,
dual_pass
,
// final boolean dual_pass,
trim_grow_pix
,
// final int trim_grow, // 3*transform_size?
width
);
// final int width) {
final
boolean
[][]
unfiltered_alpha
=
(
dbg_prefix
!=
null
)?
new
boolean
[
num_slices
][]
:
null
;
if
(
dbg_prefix
!=
null
)
{
for
(
int
i
=
0
;
i
<
num_slices
;
i
++)
{
...
...
@@ -5304,16 +5473,61 @@ public class TexturedModel {
final
double
[]
half_pix
=
new
double
[
img_size
];
final
double
[]
stitch_trim_pix
=
new
double
[
img_size
];
final
double
[]
trim_seed_pix
=
new
double
[
img_size
];
final
double
[]
seed_trim_grow_pix
=
new
double
[
img_size
];
final
double
[]
unfilt_filt_pix
=
new
double
[
img_size
];
final
double
[]
weak_fg_pix
=
new
double
[
img_size
];
final
double
[]
trim_tiles_pix
=
new
double
[
img_size
];
final
double
[]
fix_bg_pix
=
new
double
[
img_size
];
final
double
[]
fix_same_pix
=
new
double
[
img_size
];
final
double
[]
trim_alpha_pix
=
new
double
[
img_size
];
final
double
[]
grad_abs_over_same
=
new
double
[
img_size
];
final
double
[]
ridges_pix
=
new
double
[
img_size
];
final
double
[]
ridges2_pix
=
new
double
[
img_size
];
final
TileNeibs
pn
=
new
TileNeibs
(
width
,
img_size
/
width
);
for
(
int
i
=
0
;
i
<
img_size
;
i
++)
{
grad_abs_over_same
[
i
]
=
vars
[
4
][
nslice
][
i
]
/
(
vars
[
0
][
nslice
][
i
]+
seed_same_fz
);
}
final
double
min_over1
=
0.0
;
final
double
max_rel_slope
=
0.05
;
// ) { // maximal ridge slope relative to the value if >0
boolean
[]
ridges
=
pn
.
getRidges
(
grad_abs_over_same
,
// final double [] value,
null
,
// trim_pixels [nslice], // final boolean [] en,
min_over1
,
// final double min_over)
max_rel_slope
);
// final double max_rel_slope) // maximal ridge slope relative to the value if >0
final
double
min_over2
=
0.0
;
boolean
[]
ridges2
=
pn
.
getRidges
(
vars
[
4
][
nslice
],
// final double [] value,
null
,
// trim_pixels [nslice], // final boolean [] en,
min_over2
,
// final double min_over)
max_rel_slope
);
// final double max_rel_slope) // maximal ridge slope relative to the value if >0
double
[]
ridges3_pix
=
pn
.
getRidgeValue
(
vars
[
4
][
nslice
],
// final double [] value,
null
,
// final boolean [] en,
var_radius
);
// final double var_radius) now 3.5
double
[]
ridges4_pix
=
pn
.
getRidgeValue
(
grad_abs_over_same
,
// final double [] value,
null
,
// final boolean [] en,
var_radius
);
// final double var_radius) now 3.5
for
(
int
i
=
0
;
i
<
img_size
;
i
++)
{
vars_ratio
[
i
]
=
vars
[
0
][
nslice
][
i
]/
vars
[
1
][
nslice
][
i
];
vars_fom
[
i
]
=
vars
[
1
][
nslice
][
i
]/
Math
.
sqrt
(
vars
[
0
][
nslice
][
i
]);
vars_ratio
[
i
]
=
vars
[
0
][
nslice
][
i
]/(
vars
[
1
][
nslice
][
i
]+
trim_inter_fz
);
vars_fom
[
i
]
=
vars
[
1
][
nslice
][
i
]/(
vars
[
0
][
nslice
][
i
]+
seed_same_fz
);
//vars[0][nslice]
if
(
Double
.
isNaN
(
vars_ratio
[
i
]))
vars_ratio
[
i
]
=
0
;
if
(
Double
.
isNaN
(
vars_fom
[
i
]))
vars_fom
[
i
]
=
0
;
for
(
int
k
=
0
;
k
<
vars
.
length
;
k
++)
{
if
(
Double
.
isNaN
(
vars
[
k
][
nslice
][
i
]))
vars
[
k
][
nslice
][
i
]
=
0
;
}
vars_dir_ratio
[
i
]
=
vars
[
0
][
nslice
][
i
]/
dbg_out
[
1
][
nslice
][
i
];
//vars_dir_final;
// grad_abs_over_same[i] = vars[4][nslice][i] / (vars[0][nslice][i]+seed_same_fz);
ridges_pix
[
i
]
=
(
trim_pixels
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
ridges
[
i
]?
2.0
:
0.0
);
ridges2_pix
[
i
]
=
(
trim_pixels
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
ridges2
[
i
]?
2.0
:
0.0
);
half_pix
[
i
]
=
(
has_bg_pix
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
is_fg_pix
[
nslice
][
i
]?
2.0
:
0.0
);
...
...
@@ -5323,6 +5537,12 @@ public class TexturedModel {
trim_seed_pix
[
i
]
=
(
trim_pixels
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
trim_seeds
[
nslice
][
i
]?
2.0
:
0.0
);
seed_trim_grow_pix
[
i
]
=
(
trim_seeds
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
first_trimmed_alpha
[
nslice
][
i
]?
2.0
:
0.0
)
+
// after first trimming
(
unfiltered_alpha
[
nslice
][
i
]?
4.0
:
0.0
);
// grown by var_same increase
unfilt_filt_pix
[
i
]
=
(
unfiltered_alpha
[
nslice
][
i
]?
1.0
:
0.0
)
+
(
filtered_alpha
[
nslice
][
i
]?
2.0
:
0.0
);
...
...
@@ -5346,7 +5566,21 @@ public class TexturedModel {
double
[][]
dbg_img
=
{
vars
[
0
][
nslice
],
vars
[
1
][
nslice
],
vars
[
2
][
nslice
],
vars
[
3
][
nslice
],
vars
[
4
][
nslice
],
grad_abs_over_same
,
ridges_pix
,
ridges2_pix
,
ridges3_pix
,
ridges4_pix
,
vars_ratio
,
trim_fom_pix
[
nslice
],
// normalized by blurred
fom_dbg
[
0
][
nslice
],
fom_dbg
[
1
][
nslice
],
fom_dbg
[
2
][
nslice
],
fom_dbg
[
3
][
nslice
],
vars_fom
,
dbg_out
[
0
][
nslice
],
// gdbg_vars_dir_initial[nslice],
dbg_out
[
1
][
nslice
],
// gdbg_vars_dir_final[nslice],
...
...
@@ -5357,6 +5591,7 @@ public class TexturedModel {
half_pix
,
stitch_trim_pix
,
trim_seed_pix
,
seed_trim_grow_pix
,
unfilt_filt_pix
,
weak_fg_pix
,
trim_tiles_pix
,
...
...
@@ -5395,8 +5630,21 @@ public class TexturedModel {
String
[]
dbg_titles
=
{
"VAR_SAME"
,
"VAR_INTER"
,
"VAR_RATIO"
,
"VAR_FOM"
,
"GRAD_X"
,
"GRAD_Y"
,
"GRAD_ABS"
,
"GRAD_NORM"
,
"RIDGES_NORM"
,
"RIDGES_ABS"
,
"RIDGES_ANA_NORM"
,
"RIDGES_ANA_ABS"
,
"TRIM_FOM"
,
// same/(inter+trim_inter_fz)
"TRIM_FOM_NORM"
,
// same/(inter+trim_inter_fz) normalized by blurred version
"TRIM_FOM_INI"
,
// initial fom
"TRIM_FOM_FIN"
,
// final fom
"TRIM_FOM_THRESH"
,
// var_same_thresholded
"TRIM_FOM_THRESH_BLUR"
,
// var_same_thresholded_blured
"SEED_FOM"
,
// inter/(same+seed_same_fz)
"VAR_DIR_INITIAL"
,
"VAR_DIR_FINAL"
,
"DIR_INITIAL"
,
...
...
@@ -5406,6 +5654,7 @@ public class TexturedModel {
"HALF_BG_FG"
,
"STITCH_TRIM"
,
"TRIM_SEED"
,
"SEED_TRIMMED_MORE"
,
"UNFILT_FILT"
,
"WEAK_FG"
,
"TRIM_TILES"
,
...
...
src/main/java/com/elphel/imagej/tileprocessor/TileNeibs.java
View file @
ab8732b9
...
...
@@ -473,6 +473,336 @@ public class TileNeibs{
return
etiles
;
}
public
boolean
[]
getRidges
(
final
double
[]
value
,
final
boolean
[]
en
,
final
double
min_over
,
// minimal center above each side
final
double
max_rel_slope
)
{
// maximal ridge slope relative to the value if >0
final
boolean
[]
ridges
=
new
boolean
[
sizeX
*
sizeY
];
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
value
.
length
;
pix
=
ai
.
getAndIncrement
())
if
((
en
==
null
)
||
en
[
pix
]){
double
d
=
value
[
pix
];
for
(
int
dir
=
0
;
dir
<
DIRS
/
2
;
dir
++)
{
int
pix1
=
getNeibIndex
(
pix
,
dir
);
int
pix2
=
getNeibIndex
(
pix
,
(
dir
+
DIRS
/
2
)
%
DIRS
);
if
((
pix1
>=
0
)
&&
(
pix2
>=
0
))
{
if
((
d
>
value
[
pix1
])
&&
(
d
>
value
[
pix2
])
&&
((
d
-
(
value
[
pix1
]+
value
[
pix1
])/
2
)
>
min_over
))
{
// check ridge slope in ortho direction
if
(
max_rel_slope
>=
0
)
{
int
dir_ridge
=
(
dir
+
DIRS
/
4
)
%
DIRS
;
int
pix_ridge1
=
getNeibIndex
(
pix
,
dir_ridge
);
int
pix_ridge2
=
getNeibIndex
(
pix
,
(
dir_ridge
+
DIRS
/
2
)
%
DIRS
);
if
((
pix_ridge1
>=
0
)
&&
(
pix_ridge2
>=
0
))
{
double
ridge_slope
=
Math
.
abs
(
value
[
pix_ridge2
]
-
value
[
pix_ridge1
])
/
2
;
if
((
ridge_slope
/
d
)
<
max_rel_slope
)
{
ridges
[
pix
]
=
true
;
}
}
}
else
{
ridges
[
pix
]
=
true
;
break
;
}
break
;
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
return
ridges
;
}
// Not working
public
double
[]
getRidgeValue
(
final
double
[]
value
,
final
boolean
[]
en
,
final
double
var_radius
)
{
final
double
[]
ridges
=
new
double
[
sizeX
*
sizeY
];
final
int
ivar_radius
=
(
int
)
Math
.
floor
(
var_radius
);
final
double
[][]
var_weights
=
new
double
[
ivar_radius
+
1
][
ivar_radius
+
1
];
for
(
int
i
=
0
;
i
<
var_weights
.
length
;
i
++)
{
for
(
int
j
=
0
;
j
<
var_weights
[
i
].
length
;
j
++)
{
var_weights
[
i
][
j
]
=
Math
.
cos
(
0.5
*
Math
.
PI
*
i
/
var_radius
)
*
Math
.
cos
(
0.5
*
Math
.
PI
*
j
/
var_radius
);
}
}
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
Arrays
.
fill
(
ridges
,
Double
.
NaN
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
pix
=
ai
.
getAndIncrement
();
pix
<
value
.
length
;
pix
=
ai
.
getAndIncrement
())
if
((
en
==
null
)
||
en
[
pix
]){
if
(!
Double
.
isNaN
(
value
[
pix
]))
{
int
y0
=
pix
/
sizeX
;
int
x0
=
pix
%
sizeX
;
// calculate gradient
double
sw
=
0.0
,
swxd
=
0.0
,
swyd
=
0.0
;
for
(
int
dvy
=
-
ivar_radius
;
dvy
<=
ivar_radius
;
dvy
++)
{
for
(
int
dvx
=
-
ivar_radius
;
dvx
<=
ivar_radius
;
dvx
++)
{
int
indx
=
getIndex
(
x0
+
dvx
,
y0
+
dvy
);
if
(
indx
>=
0
)
{
double
d
=
value
[
indx
];
if
(!
Double
.
isNaN
(
d
))
{
double
w
=
var_weights
[
Math
.
abs
(
dvy
)][
Math
.
abs
(
dvx
)];
// 1.0;
double
wd
=
w
*
d
;
sw
+=
w
;
swxd
+=
dvx
*
wd
;
swyd
+=
dvy
*
wd
;
}
}
}
}
if
(
sw
>
0.0
)
{
// always
swxd
/=
sw
;
swyd
/=
sw
;
// normalize gradient vector
double
l
=
Math
.
sqrt
(
swxd
*
swxd
+
swyd
*
swyd
);
if
(
l
>
0
)
{
double
ux
=
swxd
/
l
;
double
uy
=
swyd
/
l
;
double
sdt2w
=
0
,
st4w
=
0
;
for
(
int
dvy
=
-
ivar_radius
;
dvy
<=
ivar_radius
;
dvy
++)
{
for
(
int
dvx
=
-
ivar_radius
;
dvx
<=
ivar_radius
;
dvx
++)
{
int
indx
=
getIndex
(
x0
+
dvx
,
y0
+
dvy
);
if
(
indx
>=
0
)
{
double
d
=
value
[
indx
];
if
(!
Double
.
isNaN
(
d
))
{
double
w
=
var_weights
[
Math
.
abs
(
dvy
)][
Math
.
abs
(
dvx
)];
// 1.0;
double
t
=
dvx
*
uy
-
dvy
*
ux
;
double
t2
=
t
*
t
;
double
t2w
=
t2
*
w
;
sdt2w
+=
d
*
t2w
;
st4w
+=
t2
*
t2w
;
}
}
}
}
if
(
st4w
>
0
)
{
ridges
[
pix
]
=
-
sdt2w
/
st4w
;
}
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
return
ridges
;
}
public
void
growSelectionGradient
(
// multithreaded version
int
grow
,
// grow tile selection by 1 over non-background tiles 1: 4 directions, 2 - 8 directions, 3 - 8 by 1, 4 by 1 more
final
boolean
[]
tiles
,
final
boolean
[]
prohibit
,
final
double
[]
value
,
final
double
min_incr
,
final
boolean
keep_top
,
// do not grow over local max
final
boolean
keep_thin_top
)
// do not grow over local max only if next after that is already selected
{
final
Thread
[]
threads
=
ImageDtt
.
newThreadArray
(
THREADS_MAX
);
final
AtomicInteger
ai
=
new
AtomicInteger
(
0
);
final
AtomicInteger
anew
=
new
AtomicInteger
(
0
);
final
boolean
[]
src_tiles
=
tiles
.
clone
();
// just in case
final
int
sizeXm1
=
sizeX
-
1
;
final
int
sizeYm1
=
sizeY
-
1
;
final
int
sizeXm2
=
sizeX
-
2
;
final
int
sizeYm2
=
sizeY
-
2
;
// grow
boolean
hor
=
true
;
final
int
dbg_tile
=
-
82228
;
// 71992; //312/112 or 61800 for 360/96
int
num_prev
=
1
;
// as if previous pass was successful
for
(;
grow
>
0
;
grow
--){
boolean
single
=
(
grow
==
1
)
&&
hor
;
System
.
arraycopy
(
tiles
,
0
,
src_tiles
,
0
,
tiles
.
length
);
anew
.
set
(
0
);
if
(
hor
){
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
tindx
=
ai
.
getAndIncrement
();
tindx
<
tiles
.
length
;
tindx
=
ai
.
getAndIncrement
())
{
int
tileX
=
tindx
%
sizeX
;
if
((
tindx
==
dbg_tile
)
||
(
tindx
==
(
dbg_tile
-
1
))){
System
.
out
.
println
(
"growSelectionMulti().1: tindx="
+
tindx
);
}
int
tindx1
=
tindx
+
1
;
if
((
tileX
<
sizeXm1
)
&&
((
prohibit
==
null
)
||
(!
prohibit
[
tindx
]
&&
!
prohibit
[
tindx1
])))
{
if
(!
src_tiles
[
tindx1
]
&&
src_tiles
[
tindx
]){
if
(
value
[
tindx1
]
>=
value
[
tindx
]
+
min_incr
)
{
boolean
add_tile
=
false
;
if
(
keep_thin_top
||
keep_top
)
{
int
tindx2
=
tindx1
+
1
;
if
((
tileX
<
sizeXm2
)
&&
!
prohibit
[
tindx2
]
&&
(
value
[
tindx2
]
<
value
[
tindx1
]))
{
if
(!
keep_top
&&
!
tiles
[
tindx2
])
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
if
(
add_tile
)
{
anew
.
getAndIncrement
();
tiles
[
tindx1
]
=
true
;
}
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
System
.
arraycopy
(
tiles
,
0
,
src_tiles
,
0
,
tiles
.
length
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
tindx
=
ai
.
getAndIncrement
();
tindx
<
tiles
.
length
;
tindx
=
ai
.
getAndIncrement
())
{
int
tileX
=
tindx
%
sizeX
;
if
((
tindx
==
dbg_tile
)
||
(
tindx
==
(
dbg_tile
+
1
))){
System
.
out
.
println
(
"growSelectionMulti().1: tindx="
+
tindx
);
}
int
tindx1
=
tindx
-
1
;
if
((
tileX
>
0
)
&&
((
prohibit
==
null
)
||
(!
prohibit
[
tindx
]
&&
!
prohibit
[
tindx1
])))
{
if
(!
src_tiles
[
tindx1
]
&&
src_tiles
[
tindx
]){
if
(
value
[
tindx1
]
>=
value
[
tindx
]
+
min_incr
)
{
boolean
add_tile
=
false
;
if
(
keep_thin_top
||
keep_top
)
{
int
tindx2
=
tindx1
-
1
;
if
((
tileX
>
1
)
&&
!
prohibit
[
tindx2
]
&&
(
value
[
tindx2
]
<
value
[
tindx1
]))
{
if
(!
keep_top
&&
!
tiles
[
tindx2
])
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
if
(
add_tile
)
{
anew
.
getAndIncrement
();
tiles
[
tindx1
]
=
true
;
}
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
}
if
(!
hor
||
single
){
// do vertically, but from previous state
System
.
arraycopy
(
tiles
,
0
,
src_tiles
,
0
,
tiles
.
length
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
tindx
=
ai
.
getAndIncrement
();
tindx
<
tiles
.
length
;
tindx
=
ai
.
getAndIncrement
())
{
int
tileY
=
tindx
/
sizeX
;
if
((
tindx
==
dbg_tile
)
||
(
tindx
==
(
dbg_tile
-
sizeX
))){
System
.
out
.
println
(
"growSelectionMulti().3: tindx="
+
tindx
);
}
int
tindx1
=
tindx
+
sizeX
;
if
((
tileY
<
sizeYm1
)
&&
((
prohibit
==
null
)
||
(!
prohibit
[
tindx
]
&&
!
prohibit
[
tindx1
])))
{
if
(!
src_tiles
[
tindx1
]
&&
src_tiles
[
tindx
]){
if
(
value
[
tindx1
]
>=
value
[
tindx
]
+
min_incr
)
{
boolean
add_tile
=
false
;
if
(
keep_thin_top
||
keep_top
)
{
int
tindx2
=
tindx1
+
sizeX
;
if
((
tileY
<
sizeYm2
)
&&
!
prohibit
[
tindx2
]
&&
(
value
[
tindx2
]
<
value
[
tindx1
]))
{
if
(!
keep_top
&&
!
tiles
[
tindx2
])
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
if
(
add_tile
)
{
anew
.
getAndIncrement
();
tiles
[
tindx1
]
=
true
;
}
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
System
.
arraycopy
(
tiles
,
0
,
src_tiles
,
0
,
tiles
.
length
);
ai
.
set
(
0
);
for
(
int
ithread
=
0
;
ithread
<
threads
.
length
;
ithread
++)
{
threads
[
ithread
]
=
new
Thread
()
{
public
void
run
()
{
for
(
int
tindx
=
ai
.
getAndIncrement
();
tindx
<
tiles
.
length
;
tindx
=
ai
.
getAndIncrement
())
{
int
tileY
=
tindx
/
sizeX
;
if
((
tindx
==
dbg_tile
)
||
(
tindx
==
(
dbg_tile
+
sizeX
))){
System
.
out
.
println
(
"growSelectionMulti().1: tindx="
+
tindx
);
}
int
tindx1
=
tindx
-
sizeX
;
if
((
tileY
>
0
)
&&
((
prohibit
==
null
)
||
(!
prohibit
[
tindx
]
&&
!
prohibit
[
tindx1
])))
{
if
(!
src_tiles
[
tindx1
]
&&
src_tiles
[
tindx
]){
if
(
value
[
tindx1
]
>=
value
[
tindx
]
+
min_incr
)
{
boolean
add_tile
=
false
;
if
(
keep_thin_top
||
keep_top
)
{
int
tindx2
=
tindx1
-
sizeX
;
if
((
tileY
>
1
)
&&
!
prohibit
[
tindx2
]
&&
(
value
[
tindx2
]
<
value
[
tindx1
]))
{
if
(!
keep_top
&&
!
tiles
[
tindx2
])
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
}
else
{
add_tile
=
true
;
}
if
(
add_tile
)
{
anew
.
getAndIncrement
();
tiles
[
tindx1
]
=
true
;
}
}
}
}
}
}
};
}
ImageDtt
.
startAndJoin
(
threads
);
}
hor
=
!
hor
;
if
((
anew
.
get
()
==
0
)
&&
(
num_prev
==
0
)){
break
;
}
num_prev
=
anew
.
get
();
}
}
public
void
growSelectionMulti
(
// multithreaded version
int
grow
,
// grow tile selection by 1 over non-background tiles 1: 4 directions, 2 - 8 directions, 3 - 8 by 1, 4 by 1 more
final
boolean
[]
tiles
,
...
...
@@ -587,6 +917,9 @@ public class TileNeibs{
}
}
public
static
boolean
isMainThread
()
{
return
Thread
.
currentThread
().
getName
().
equals
(
"main"
);
}
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment