I created a script that automatically closes small gaps and performs continuous area selection. I made it with the help of many people on this site. Thanks to all of you.
<test picture
<Fill in the resulting selection
You can try it out by setting up this script in Ten Scripts and running the script at any cursor position. If it works correctly, you can make the selections shown in the image above. (coloring is not included in the code)
from struct import iter_unpack
from krita import *
from PyQt5.QtCore import (
Qt,
QPoint,
QPointF,
QTimer,
QEvent)
from PyQt5.QtGui import (
QImage,
QPainter,
QColor,
QTabletEvent,
QCursor)
from PyQt5.QtWidgets import (
QWidget,
QApplication)
import time
def find_current_canvas():
q_window = application.activeWindow().qwindow()
q_stacked_widget = q_window.centralWidget()
q_mdi_area = q_stacked_widget.currentWidget()
q_mdi_sub_window = q_mdi_area.currentSubWindow()
view = q_mdi_sub_window.widget()
for c in view.children():
if c.metaObject().className() == 'KisCanvasController':
viewport = c.viewport()
canvas = viewport.findChild(QWidget)
return canvas
def click_canvas(global_pos=None):#click action
if global_pos is None:
global_pos = pos#これが現在のカーソル位置
global_posF = QPointF(global_pos)
posF = QPointF(canvas.mapFromGlobal(global_pos))
else:
global_pos = pos
global_posF = QPointF(global_pos)
posF = QPointF(canvas.mapFromGlobal(global_pos))
device = QTabletEvent.Stylus
pointer_type = QTabletEvent.Pen
pressure = 1
x_tilt = 0
y_tilt = 0
tangential_pressure = 0.0
rotation = 0.0
z_pos = 0
key_state = Qt.NoModifier
unique_id = 1234
button = Qt.LeftButton
buttons = Qt.LeftButton
table_press = QTabletEvent(
QEvent.TabletPress,
posF,
global_posF,
device,
pointer_type,
pressure,
x_tilt,
y_tilt,
tangential_pressure,
rotation,
z_pos,
key_state,
unique_id,
button,
buttons)
table_release = QTabletEvent(
QEvent.TabletRelease,
posF,
global_posF,
device,
pointer_type,
pressure,
x_tilt,
y_tilt,
tangential_pressure,
rotation,
z_pos,
key_state,
unique_id,
button,
buttons)
if not canvas.isActiveWindow():
canvas.activateWindow()
QApplication.sendEvent(canvas, table_press)
QApplication.sendEvent(canvas, table_release)
activeDoc.waitForDone()
def edgeDetectFilter(node):
filt = application.filter('edge detection')
configParameters = filt.configuration()
configParameters.setProperties({
'horizRadius': 1,
'lockAspect': True,
'output': 'pythagorean',
'transparency': True,
'type': 'prewitt',
'vertRadius': 1})
filt.setConfiguration(configParameters)
filt.apply(node, 0, 0, w, h)
activeDoc.waitForDone()
def createSelection(node, filterFunc):#Converting an image to selection
bounds = node.bounds()
x, y = bounds.left(), bounds.top()
width, height = bounds.width(), bounds.height()
#プロジェクションを使ってるから画面に出さないといけない
pixelData = bytes(node.projectionPixelData(x, y, width, height))
selectionData = bytearray(width * height)
for i, rgba in enumerate(iter_unpack('>BBBB', pixelData)):
selectionData[i] = filterFunc(*rgba)
newSelection = Selection()
newSelection.setPixelData(selectionData, x, y, width, height)
return newSelection
def transparentFilter(r, g, b, a):#不透明度の閾値を数値で決める
return 255 if a > 1 else 0#aが何より大きかったら完全に選択するか
#使い方selection = createSelection(doc.activeNode(), transparentFilter)
def fillColor(node, colorx):#選択範囲に関係なく全面が色づく
image = QImage(w, h, QImage.Format_ARGB32)
image.fill(colorx)
pixel_ptr = image.constBits()
pixels = bytes(pixel_ptr.asarray(image.byteCount()))
node.setPixelData(pixels, 0, 0, w, h)
activeDoc.waitForDone()
def makeProcessedLine():
#Make the entire image a line drawing
lineNode = rootCopy.duplicate()
edgeDetectFilter(lineNode)#エッジを出す
#To Selection
ProcessedLineSe = createSelection(lineNode, transparentFilter)
return ProcessedLineSe
def reactsToGap():
#Scope to fill in the gap location
reactsToGapSe = processedLineSe.duplicate()
eraseLineSe = processedLineSe.duplicate()
reactsToGapSe.grow(1, 1)#この辺で調整
reactsToGapSe.grow(1, 1)
reactsToGapSe.grow(1, 1)
activeDoc.setSelection(reactsToGapSe)
reactsToGapSe.shrink(4, 4, True)
reactsToGapSe.subtract(eraseLineSe)
reactsToGapSe.grow(2, 2)
return reactsToGapSe
def exclusionZone():
#Scope for eliminating unwanted parts of the lid.
selectionPL1 = processedLineSe.duplicate()
selectionPL2 = processedLineSe.duplicate()
selectionPL3 = processedLineSe.duplicate()
#Apex
activeDoc.setSelection(selectionPL1)
selectionPL1.grow(3, 3)
selectionPL1.shrink(6, 6, True)
activeDoc.setSelection(selectionPL2)
selectionPL2.grow(7, 7)#このへんの数字はいじりようがある
selectionPL2.shrink(3, 3, True)
selectionPL3.grow(1, 1)
selectionPL2.subtract(selectionPL3)
selectionPL1.add(selectionPL2)
return selectionPL1
def fillExpantion():#Expand the selection inside the line
s = activeDoc.selection()
s2 = s.duplicate()
for i in range(6):
s2.subtract(processedLineSe)
s2.grow(1, 1)
s2.intersect(reactsToGapSe)#蓋と拡大する範囲の交錯部分をプラス
s2.intersect(allSe)
s.add(s2)
activeDoc.setSelection(s)
def makeCloser():
#赤いレイヤを用意 clip red layer
redNode = activeDoc.createNode("redNode", "paintlayer")
fillColor(redNode, Qt.red)
allSe.copy(redNode)
closerNode = activeDoc.createNode("closerNode", "paintlayer")
closerNode.setBlendingMode("not_converse")#否定逆論理
processedLineSe2 = processedLineSe.duplicate()
activeDoc.setSelection(processedLineSe2)#最も狭い隙間を閉じるもの
processedLineSe2.grow(1, 1)
processedLineSe2.shrink(2, 2, True)
reactsToGapSe.grow(1, 1)
reactsToGapSe.add(processedLineSe2)
reactsToGapSe.subtract(exSe)#フタ完成
reactsToGapSe.paste(closerNode, 0, 0)#クローザーを設置
rootNode.addChildNode(closerNode, topNode)
application.action('deselect').trigger()#変な選択あるから一旦掃除
activeDoc.waitForDone()
click_canvas()
fillExpantion()
closerNode.remove()
application = Krita.instance()
activeDoc = application.activeDocument()
activeLayer = activeDoc.activeNode()
rootNode = activeDoc.rootNode()
topNode = activeDoc.topLevelNodes().pop()
w = activeDoc.width()
h = activeDoc.height()
allSe = Selection()
allSe.select(0, 0, w, h, 255)
# time.sleep(2)
pos = QCursor.pos()#起動時点のカーソル位置
#Avoid having the same tool twice in a row.
application.action('PanTool').trigger()#連続で同じツールを持たないよう回避
application.action('KisToolSelectContiguous').trigger()#select tool
rootCopy = activeDoc.createNode("rootCopy", "paintlayer")
rootNodepix = rootNode.projectionPixelData(0, 0, w, h)
rootCopy.setPixelData(rootNodepix,0, 0, w, h)#ノード全体に画像貼り付け
canvas = find_current_canvas()
processedLineSe = makeProcessedLine()
reactsToGapSe = reactsToGap()
exSe = exclusionZone()
makeCloser()
activeDoc.setActiveNode(activeLayer)
#Make an ant march appear.
application.action('invert_selection').trigger()#アリの行進
application.action('invert_selection').trigger()
(If you enable time.sleep(2) on line 246, you can also use it in the scripter, since execution will be after two seconds.)
However, it is not yet a practical script for the following reasons
Current problems and suggestions for improvement
- Slow operation when targeting images larger than 1000Pix.
- It does not respond when the active layer is a group layer.Why does it work when used in Scripter…?
- There are small areas that fail to be painted.
- Can’t I use shrink() for selections without activeDoc.setSelection()?
- Is there a way to use continuous selection without click action? If there is another, the code might run faster.
I will continue to tweak and improve this script. If anyone has ideas on how to make this better, I would appreciate it if you would share them with me. I am a programming beginner so if you can share your code with me, that would be great.
Translated at DeepL
I will go test this later tonight.
I don’t think there’s any documentation on how things work in Krita, so the painstaking trial and error seems like the only way to find out. Of course looking at the change history for other fill-related patches helps a lot.
