diff --git a/libs/image/kis_edge_detection_kernel.cpp b/libs/image/kis_edge_detection_kernel.cpp
index 7872414e08..073f7efe5c 100644
--- a/libs/image/kis_edge_detection_kernel.cpp
+++ b/libs/image/kis_edge_detection_kernel.cpp
@@ -1,651 +1,667 @@
 /*
  *  Copyright (c) 2017 Wolthera van Hövell tot Westerflier <griffinvalley@gmail.com>
  *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
  *  the Free Software Foundation; either version 2 of the License, or
  *  (at your option) any later version.
  *
  *  This program is distributed in the hope that it will be useful,
  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  *  GNU General Public License for more details.
  *
  *  You should have received a copy of the GNU General Public License
  *  along with this program; if not, write to the Free Software
  *  Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  */
 
 #include "kis_edge_detection_kernel.h"
 #include "kis_global.h"
 #include "kis_convolution_kernel.h"
 #include <kis_convolution_painter.h>
 #include <KoCompositeOpRegistry.h>
 #include <QRect>
 #include <KoColorSpace.h>
 #include <kis_iterator_ng.h>
 #include <QVector3D>
 #include <kis_random_accessor_ng.h>
 
 KisEdgeDetectionKernel::KisEdgeDetectionKernel()
 {
 
 }
 /*
  * This code is very similar to the gaussian kernel code, except unlike the gaussian code,
  * edge-detection kernels DO use the diagonals.
  * Except for the simple mode. We implement the simple mode because it is an analog to
  * the old sobel filter.
  */
 
 Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> KisEdgeDetectionKernel::createHorizontalMatrix(qreal radius,
                                                                                                     FilterType type,
                                                                                                     bool reverse)
 {
     int kernelSize = kernelSizeFromRadius(radius);
     Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> matrix(kernelSize, kernelSize);
 
     KIS_ASSERT_RECOVER_NOOP(kernelSize & 0x1);
     const int center = kernelSize / 2;
 
     if (type==Prewit) {
         for (int x = 0; x < kernelSize; x++) {
             for (int y=0; y<kernelSize; y++) {
                 qreal xDistance;
                 if (reverse) {
                     xDistance = x - center;
                 } else {
                     xDistance = center - x;
                 }
                 matrix(x, y) = xDistance;
             }
         }
     } else if(type==Simple) {
         matrix.resize(kernelSize, 1);
         for (int x = 0; x < kernelSize; x++) {
             qreal xDistance;
             if (reverse) {
                 xDistance = x - center;
             } else {
                 xDistance = center - x;
             }
             if (x==center) {
                 matrix(x, 0) = 0;
             } else {
                 matrix(x, 0) = (1/xDistance);
             }
         }
     } else {
         for (int x = 0; x < kernelSize; x++) {
             for (int y=0; y<kernelSize; y++) {
                 if (x==center && y==center) {
                     matrix(x, y) = 0;
                 } else {
                     qreal xD, yD;
                     if (reverse) {
                         xD = x - center;
                         yD = y - center;
                     } else {
                         xD = center - x;
                         yD = center - y;
                     }
                     matrix(x, y) = xD / (xD*xD + yD*yD);
                 }
             }
         }
     }
     return matrix;
 }
 
 Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> KisEdgeDetectionKernel::createVerticalMatrix(qreal radius,
                                                                                                   FilterType type,
                                                                                                   bool reverse)
 {
     int kernelSize = kernelSizeFromRadius(radius);
 
     Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> matrix(kernelSize, kernelSize);
     KIS_ASSERT_RECOVER_NOOP(kernelSize & 0x1);
     const int center = kernelSize / 2;
 
     if (type==Prewit) {
         for (int y = 0; y < kernelSize; y++) {
             for (int x=0; x<kernelSize; x++) {
                 qreal yDistance;
                 if (reverse) {
                     yDistance = y - center;
                 } else {
                     yDistance = center - y;
                 }
                 matrix(x, y) = yDistance;
             }
         }
     } else if(type==Simple) {
         matrix.resize(1, kernelSize);
         for (int y = 0; y < kernelSize; y++) {
             qreal yDistance;
             if (reverse) {
                 yDistance = y - center;
             } else {
                 yDistance = center - y;
             }
             if (y==center) {
                 matrix(0, y) = 0;
             } else {
                 matrix(0, y) = (1/yDistance);
             }
         }
     } else {
         for (int y = 0; y < kernelSize; y++) {
             for (int x=0; x<kernelSize; x++) {
                 if (x==center && y==center) {
                     matrix(x, y) = 0;
                 } else {
                     qreal xD, yD;
                     if (reverse) {
                         xD = x - center;
                         yD = y - center;
                     } else {
                         xD = center - x;
                         yD = center - y;
                     }
                     matrix(x, y) = yD / (xD*xD + yD*yD);
                 }
             }
         }
     }
     return matrix;
 }
 
 KisConvolutionKernelSP KisEdgeDetectionKernel::createHorizontalKernel(qreal radius,
                                                                       KisEdgeDetectionKernel::FilterType type,
                                                                       bool denormalize,
                                                                       bool reverse)
 {
     Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> matrix = createHorizontalMatrix(radius, type, reverse);
     if (denormalize) {
         return KisConvolutionKernel::fromMatrix(matrix, 0.5, 1);
     } else {
         return KisConvolutionKernel::fromMatrix(matrix, 0, matrix.sum());
     }
 }
 
 KisConvolutionKernelSP KisEdgeDetectionKernel::createVerticalKernel(qreal radius,
                                                                     KisEdgeDetectionKernel::FilterType type,
                                                                     bool denormalize,
                                                                     bool reverse)
 {
     Eigen::Matrix<qreal, Eigen::Dynamic, Eigen::Dynamic> matrix = createVerticalMatrix(radius, type, reverse);
     if (denormalize) {
         return KisConvolutionKernel::fromMatrix(matrix, 0.5, 1);
     } else {
         return KisConvolutionKernel::fromMatrix(matrix, 0, matrix.sum());
     }
 }
 
 int KisEdgeDetectionKernel::kernelSizeFromRadius(qreal radius)
 {
     return qMax((int)(2 * ceil(sigmaFromRadius(radius)) + 1), 3);
 }
 
 qreal KisEdgeDetectionKernel::sigmaFromRadius(qreal radius)
 {
     return 0.3 * radius + 0.3;
 }
 
 void KisEdgeDetectionKernel::applyEdgeDetection(KisPaintDeviceSP device,
                                                 const QRect &rect,
                                                 qreal xRadius,
                                                 qreal yRadius,
                                                 KisEdgeDetectionKernel::FilterType type,
                                                 const QBitArray &channelFlags,
                                                 KoUpdater *progressUpdater,
                                                 FilterOutput output,
                                                 bool writeToAlpha)
 {
     QPoint srcTopLeft = rect.topLeft();
     KisPainter finalPainter(device);
     finalPainter.setChannelFlags(channelFlags);
     finalPainter.setProgress(progressUpdater);
     if (output == pythagorean || output == radian) {
         KisPaintDeviceSP x_denormalised = new KisPaintDevice(device->colorSpace());
         KisPaintDeviceSP y_denormalised = new KisPaintDevice(device->colorSpace());
 
         x_denormalised->prepareClone(device);
         y_denormalised->prepareClone(device);
 
 
         KisConvolutionKernelSP kernelHorizLeftRight = KisEdgeDetectionKernel::createHorizontalKernel(xRadius, type);
         KisConvolutionKernelSP kernelVerticalTopBottom = KisEdgeDetectionKernel::createVerticalKernel(yRadius, type);
 
         qreal horizontalCenter = qreal(kernelHorizLeftRight->width()) / 2.0;
         qreal verticalCenter = qreal(kernelVerticalTopBottom->height()) / 2.0;
 
         KisConvolutionPainter horizPainterLR(x_denormalised);
         horizPainterLR.setChannelFlags(channelFlags);
         horizPainterLR.setProgress(progressUpdater);
         horizPainterLR.applyMatrix(kernelHorizLeftRight, device,
                                    srcTopLeft - QPoint(0, ceil(horizontalCenter)),
                                    srcTopLeft - QPoint(0, ceil(horizontalCenter)),
                                    rect.size() + QSize(0, 2 * ceil(horizontalCenter)), BORDER_REPEAT);
 
 
         KisConvolutionPainter verticalPainterTB(y_denormalised);
         verticalPainterTB.setChannelFlags(channelFlags);
         verticalPainterTB.setProgress(progressUpdater);
         verticalPainterTB.applyMatrix(kernelVerticalTopBottom, device,
                                       srcTopLeft - QPoint(0, ceil(verticalCenter)),
                                       srcTopLeft - QPoint(0, ceil(verticalCenter)),
                                       rect.size() + QSize(0, 2 * ceil(verticalCenter)), BORDER_REPEAT);
 
         KisSequentialIterator yItterator(y_denormalised, rect);
         KisSequentialIterator xItterator(x_denormalised, rect);
         KisSequentialIterator finalIt(device, rect);
         const int pixelSize = device->colorSpace()->pixelSize();
         const int channels = device->colorSpace()->channelCount();
         const int alphaPos = device->colorSpace()->alphaPos();
         KIS_SAFE_ASSERT_RECOVER_RETURN(alphaPos >= 0);
 
         QVector<float> yNormalised(channels);
         QVector<float> xNormalised(channels);
         QVector<float> finalNorm(channels);
 
         while(yItterator.nextPixel() && xItterator.nextPixel() && finalIt.nextPixel()) {
             device->colorSpace()->normalisedChannelsValue(yItterator.rawData(), yNormalised);
             device->colorSpace()->normalisedChannelsValue(xItterator.rawData(), xNormalised);
             device->colorSpace()->normalisedChannelsValue(finalIt.rawData(), finalNorm);
 
             if (output == pythagorean) {
                 for (int c = 0; c<channels; c++) {
                     finalNorm[c] = 2 * sqrt( ((xNormalised[c]-0.5)*(xNormalised[c]-0.5)) + ((yNormalised[c]-0.5)*(yNormalised[c]-0.5)));
                 }
             } else { //radian
                 for (int c = 0; c<channels; c++) {
                     finalNorm[c] = atan2(xNormalised[c]-0.5, yNormalised[c]-0.5);
                 }
             }
 
             if (writeToAlpha) {
                 KoColor col(finalIt.rawData(), device->colorSpace());
                 qreal alpha = 0;
 
                 for (int c = 0; c<(channels-1); c++) {
                     alpha = alpha+finalNorm[c];
                 }
 
                 alpha = qMin(alpha/(channels-1), col.opacityF());
                 col.setOpacity(alpha);
                 memcpy(finalIt.rawData(), col.data(), pixelSize);
             } else {
                 quint8* f = finalIt.rawData();
                 finalNorm[alphaPos] = 1.0;
                 device->colorSpace()->fromNormalisedChannelsValue(f, finalNorm);
                 memcpy(finalIt.rawData(), f, pixelSize);
             }
 
         }
     } else {
         KisConvolutionKernelSP kernel;
         qreal center = 0;
         bool denormalize = !writeToAlpha;
         if (output == xGrowth) {
             kernel = KisEdgeDetectionKernel::createHorizontalKernel(xRadius, type, denormalize);
             center = qreal(kernel->width()) / 2.0;
         } else if (output == xFall) {
             kernel = KisEdgeDetectionKernel::createHorizontalKernel(xRadius, type, denormalize, true);
             center = qreal(kernel->width()) / 2.0;
         } else if (output == yGrowth) {
             kernel = KisEdgeDetectionKernel::createVerticalKernel(yRadius, type, denormalize);
             center = qreal(kernel->height()) / 2.0;
         } else { //yFall
             kernel = KisEdgeDetectionKernel::createVerticalKernel(yRadius, type, denormalize, true);
             center = qreal(kernel->height()) / 2.0;
         }
 
         if (writeToAlpha) {
             KisPaintDeviceSP denormalised = new KisPaintDevice(device->colorSpace());
             denormalised->prepareClone(device);
 
             KisConvolutionPainter kernelP(denormalised);
             kernelP.setChannelFlags(channelFlags);
             kernelP.setProgress(progressUpdater);
             kernelP.applyMatrix(kernel, device,
                                 srcTopLeft - QPoint(0, ceil(center)),
                                 srcTopLeft - QPoint(0, ceil(center)),
                                 rect.size() + QSize(0, 2 * ceil(center)), BORDER_REPEAT);
             KisSequentialIterator iterator(denormalised, rect);
             KisSequentialIterator finalIt(device, rect);
             const int pixelSize = device->colorSpace()->pixelSize();
             const int channels = device->colorSpace()->colorChannelCount();
             QVector<float> normalised(channels);
             while (iterator.nextPixel() && finalIt.nextPixel()) {
                 device->colorSpace()->normalisedChannelsValue(iterator.rawData(), normalised);
                 KoColor col(finalIt.rawData(), device->colorSpace());
                 qreal alpha = 0;
                 for (int c = 0; c<channels; c++) {
                     alpha = alpha+normalised[c];
                 }
                 alpha = qMin(alpha/channels, col.opacityF());
                 col.setOpacity(alpha);
                 memcpy(finalIt.rawData(), col.data(), pixelSize);
 
             }
 
         } else {
             KisConvolutionPainter kernelP(device);
             kernelP.setChannelFlags(channelFlags);
             kernelP.setProgress(progressUpdater);
             kernelP.applyMatrix(kernel, device,
                                 srcTopLeft - QPoint(0, ceil(center)),
                                 srcTopLeft - QPoint(0, ceil(center)),
                                 rect.size() + QSize(0, 2 * ceil(center)), BORDER_REPEAT);
 
             KisSequentialIterator finalIt(device, rect);
             int numConseqPixels = finalIt.nConseqPixels();
             while (finalIt.nextPixels(numConseqPixels)) {
                 numConseqPixels = finalIt.nConseqPixels();
                 device->colorSpace()->setOpacity(finalIt.rawData(), 1.0, numConseqPixels);
             }
         }
     }
 }
 
 void KisEdgeDetectionKernel::applyCannyEdgeDetection(KisPaintDeviceSP _device,
                                                      const QRect& rect,
                                                      qreal xRadius,
                                                      qreal yRadius,
                                                      const QBitArray& channelFlags,
                                                      KoUpdater* progressUpdater,
                                                      int thresholdMax,
                                                      int thresholdMin)
 {
     QPoint srcTopLeft = rect.topLeft();
 
     auto bounds = _device->defaultBounds();
     auto device = KisPainter::convertToAlphaAsGray(_device);
     device->setDefaultBounds(bounds);
 
     KisPaintDeviceSP x_denormalised = new KisPaintDevice(device->colorSpace());
     KisPaintDeviceSP y_denormalised = new KisPaintDevice(device->colorSpace());
 
     x_denormalised->prepareClone(device);
     y_denormalised->prepareClone(device);
 
-    KisConvolutionKernelSP kernelHorizLeftRight =
-        KisEdgeDetectionKernel::createHorizontalKernel(xRadius, KisEdgeDetectionKernel::SobelVector);
-    KisConvolutionKernelSP kernelVerticalTopBottom =
-        KisEdgeDetectionKernel::createVerticalKernel(yRadius, KisEdgeDetectionKernel::SobelVector);
+    Eigen::Matrix<qreal, 3, 3> horizontalMatrix;
+    horizontalMatrix << -1, 0, 1,
+                     -2, 0, 2,
+                     -1, 0, 1;
 
-    qreal horizontalCenter = qreal(kernelHorizLeftRight->width()) / 2.0;
-    qreal verticalCenter = qreal(kernelVerticalTopBottom->height()) / 2.0;
+    auto verticalMatrix = horizontalMatrix.transpose();
+
+    auto kernelHorizLeftRight = KisConvolutionKernel::fromMatrix(horizontalMatrix, 0.5, 1);
+    auto kernelVerticalTopBottom = KisConvolutionKernel::fromMatrix(verticalMatrix, 0.5, 1);
 
     KisConvolutionPainter horizPainterLR(x_denormalised);
     horizPainterLR.setChannelFlags(channelFlags);
     horizPainterLR.setProgress(progressUpdater);
     horizPainterLR.applyMatrix(kernelHorizLeftRight,
                                device,
-                               srcTopLeft - QPoint(0, ceil(horizontalCenter)),
-                               srcTopLeft - QPoint(0, ceil(horizontalCenter)),
-                               rect.size() + QSize(0, 2 * ceil(horizontalCenter)),
+                               srcTopLeft - QPoint(0, 1),
+                               srcTopLeft - QPoint(0, 1),
+                               rect.size() + QSize(0, 2),
                                BORDER_REPEAT);
 
     KisConvolutionPainter verticalPainterTB(y_denormalised);
     verticalPainterTB.setChannelFlags(channelFlags);
     verticalPainterTB.setProgress(progressUpdater);
     verticalPainterTB.applyMatrix(kernelVerticalTopBottom,
                                   device,
-                                  srcTopLeft - QPoint(0, ceil(verticalCenter)),
-                                  srcTopLeft - QPoint(0, ceil(verticalCenter)),
-                                  rect.size() + QSize(0, 2 * ceil(verticalCenter)),
+                                  srcTopLeft - QPoint(0, 1),
+                                  srcTopLeft - QPoint(0, 1),
+                                  rect.size() + QSize(0, 2),
                                   BORDER_REPEAT);
 
     KisSequentialIterator yIterator(y_denormalised, rect);
     KisSequentialIterator xIterator(x_denormalised, rect);
     KisSequentialIterator sobelIterator(device, rect);
 
     QVector<float> yNormalised(1);
     QVector<float> xNormalised(1);
     QVector<float> finalNorm(1);
 
     while (yIterator.nextPixel() && xIterator.nextPixel() && sobelIterator.nextPixel()) {
         device->colorSpace()->normalisedChannelsValue(yIterator.rawData(), yNormalised);
         device->colorSpace()->normalisedChannelsValue(xIterator.rawData(), xNormalised);
         device->colorSpace()->normalisedChannelsValue(sobelIterator.rawData(), finalNorm);
-        finalNorm[0] = 2 * sqrt(((xNormalised[0] - 0.5) * (xNormalised[0] - 0.5)) +
-                                ((yNormalised[0] - 0.5) * (yNormalised[0] - 0.5)));
+        finalNorm[0] = 2 * hypot(xNormalised[0] - 0.5, yNormalised[0] - 0.5);
         quint8* f = sobelIterator.rawData();
         device->colorSpace()->fromNormalisedChannelsValue(f, finalNorm);
         memcpy(sobelIterator.rawData(), f, device->colorSpace()->pixelSize());
     }
 
     KisRandomAccessorSP randomSobel = device->createRandomAccessorNG(rect.x(), rect.y());
     KisRandomConstAccessorSP gx_iter = x_denormalised->createRandomConstAccessorNG(rect.x(), rect.y());
-    KisRandomConstAccessorSP gy_iter = x_denormalised->createRandomConstAccessorNG(rect.x(), rect.y());
+    KisRandomConstAccessorSP gy_iter = y_denormalised->createRandomConstAccessorNG(rect.x(), rect.y());
 
     for (int x = rect.x() + 1; x < rect.x() + rect.width() - 1; x++) {
         for (int y = rect.y() + 1; y < rect.y() + rect.height() - 1; y++) {
             randomSobel->moveTo(x, y);
-            const float current = *randomSobel->rawData();
             float firstIntensity, secondIntensity;
             gx_iter->moveTo(x, y);
             gy_iter->moveTo(x, y);
-            const float direction = atan2(*gx_iter->rawDataConst(), *gy_iter->rawDataConst());
+            device->colorSpace()->normalisedChannelsValue(gx_iter->rawDataConst(), yNormalised);
+            device->colorSpace()->normalisedChannelsValue(gy_iter->rawDataConst(), xNormalised);
+            device->colorSpace()->normalisedChannelsValue(randomSobel->rawDataConst(), finalNorm);
+
+            const float direction = static_cast<float>((fmod(atan2(yNormalised[0] - 0.5, xNormalised[0] - 0.5) + M_PI, M_PI)/M_PI) * 8);
 
             if (direction <= 1 || direction > 7) {
                 randomSobel->moveTo(x + 1, y);
-                firstIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gx_iter->rawDataConst(), yNormalised);
+                firstIntensity = yNormalised[0];
+
                 randomSobel->moveTo(x - 1, y);
-                secondIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gy_iter->rawDataConst(), xNormalised);
+                secondIntensity = xNormalised[0];
             }
 
-            if (direction > 1 || direction <= 3) {
+            if (direction > 1 && direction <= 3) {
                 randomSobel->moveTo(x + 1, y + 1);
-                firstIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gx_iter->rawDataConst(), yNormalised);
+                firstIntensity = yNormalised[0];
+
                 randomSobel->moveTo(x - 1, y - 1);
-                secondIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gy_iter->rawDataConst(), xNormalised);
+                secondIntensity = xNormalised[0];
             }
 
-            if (direction > 3 || direction <= 5) {
+            if (direction > 3 && direction <= 5) {
                 randomSobel->moveTo(x, y + 1);
-                firstIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gx_iter->rawDataConst(), yNormalised);
+                firstIntensity = yNormalised[0];
+
                 randomSobel->moveTo(x, y - 1);
-                secondIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gy_iter->rawDataConst(), xNormalised);
+                secondIntensity = xNormalised[0];
             }
 
-            if (direction > 5 || direction <= 7) {
+            if (direction > 5 && direction <= 7) {
                 randomSobel->moveTo(x + 1, y - 1);
-                firstIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gx_iter->rawDataConst(), yNormalised);
+                firstIntensity = yNormalised[0];
+
                 randomSobel->moveTo(x - 1, y + 1);
-                secondIntensity = *randomSobel->rawData();
+                device->colorSpace()->normalisedChannelsValue(gy_iter->rawDataConst(), xNormalised);
+                secondIntensity = xNormalised[0];
             }
 
             randomSobel->moveTo(x, y);
 
-            if (2 * current <= (firstIntensity + secondIntensity)) {
+            if (finalNorm[0] < firstIntensity || finalNorm[0] < secondIntensity) {
                 quint8* temp = new quint8();
-                finalNorm[0] = 0.0;
+                finalNorm[0] = 0;
                 device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                 memcpy(randomSobel->rawData(), temp, device->pixelSize());
             }
         }
     }
 
     for (int x = rect.x() + 1; x < rect.x() + rect.width() - 1; x++) {
         for (int y = rect.y() + 1; y < rect.y() + rect.height() - 1; y++) {
             randomSobel->moveTo(x, y);
             quint8 intensity = *randomSobel->rawData();
             if (intensity < thresholdMin) {
                 quint8* temp = new quint8();
                 finalNorm[0] = 0.0;
                 device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                 memcpy(randomSobel->rawData(), temp, device->pixelSize());
             } else if (intensity < thresholdMax) {
                 randomSobel->moveTo(x + 1, y);
                 quint8 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x - 1, y);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x, y + 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x, y - 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x + 1, y + 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x - 1, y - 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x + 1, y - 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
                 randomSobel->moveTo(x - 1, y + 1);
                 intensity = *randomSobel->rawData();
                 if (intensity > thresholdMax) {
                     quint8* temp = new quint8();
                     finalNorm[0] = 1.0;
                     device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                     memcpy(randomSobel->rawData(), temp, device->pixelSize());
                     continue;
                 }
 
             } else {
                 quint8* temp = new quint8();
                 finalNorm[0] = 1.0;
                 device->colorSpace()->fromNormalisedChannelsValue(temp, finalNorm);
                 memcpy(randomSobel->rawData(), temp, device->pixelSize());
             }
         }
     }
 
     KisPainter::copyAreaOptimized(rect.topLeft(), device, _device, rect);
 }
 
 void KisEdgeDetectionKernel::convertToNormalMap(KisPaintDeviceSP device,
                                                 const QRect &rect,
                                                 qreal xRadius,
                                                 qreal yRadius,
                                                 KisEdgeDetectionKernel::FilterType type,
                                                 int channelToConvert,
                                                 QVector<int> channelOrder,
                                                 QVector<bool> channelFlip,
                                                 const QBitArray &channelFlags,
                                                 KoUpdater *progressUpdater)
 {
     QPoint srcTopLeft = rect.topLeft();
     KisPainter finalPainter(device);
     finalPainter.setChannelFlags(channelFlags);
     finalPainter.setProgress(progressUpdater);
     KisPaintDeviceSP x_denormalised = new KisPaintDevice(device->colorSpace());
     KisPaintDeviceSP y_denormalised = new KisPaintDevice(device->colorSpace());
     x_denormalised->prepareClone(device);
     y_denormalised->prepareClone(device);
 
     KisConvolutionKernelSP kernelHorizLeftRight = KisEdgeDetectionKernel::createHorizontalKernel(yRadius, type, true, !channelFlip[1]);
     KisConvolutionKernelSP kernelVerticalTopBottom = KisEdgeDetectionKernel::createVerticalKernel(xRadius, type, true, !channelFlip[0]);
 
     qreal horizontalCenter = qreal(kernelHorizLeftRight->width()) / 2.0;
     qreal verticalCenter = qreal(kernelVerticalTopBottom->height()) / 2.0;
 
     KisConvolutionPainter horizPainterLR(y_denormalised);
     horizPainterLR.setChannelFlags(channelFlags);
     horizPainterLR.setProgress(progressUpdater);
     horizPainterLR.applyMatrix(kernelHorizLeftRight, device,
                                srcTopLeft - QPoint(ceil(horizontalCenter), 0),
                                srcTopLeft - QPoint(ceil(horizontalCenter), 0),
                                rect.size() + QSize(2 * ceil(horizontalCenter), 0), BORDER_REPEAT);
 
 
     KisConvolutionPainter verticalPainterTB(x_denormalised);
     verticalPainterTB.setChannelFlags(channelFlags);
     verticalPainterTB.setProgress(progressUpdater);
     verticalPainterTB.applyMatrix(kernelVerticalTopBottom, device,
                                   srcTopLeft - QPoint(0, ceil(verticalCenter)),
                                   srcTopLeft - QPoint(0, ceil(verticalCenter)),
                                   rect.size() + QSize(0, 2 * ceil(verticalCenter)), BORDER_REPEAT);
 
     KisSequentialIterator yItterator(y_denormalised, rect);
     KisSequentialIterator xItterator(x_denormalised, rect);
     KisSequentialIterator finalIt(device, rect);
     const int pixelSize = device->colorSpace()->pixelSize();
     const int channels = device->colorSpace()->channelCount();
     const int alphaPos = device->colorSpace()->alphaPos();
     KIS_SAFE_ASSERT_RECOVER_RETURN(alphaPos >= 0);
 
     QVector<float> yNormalised(channels);
     QVector<float> xNormalised(channels);
     QVector<float> finalNorm(channels);
 
     while(yItterator.nextPixel() && xItterator.nextPixel() && finalIt.nextPixel()) {
         device->colorSpace()->normalisedChannelsValue(yItterator.rawData(), yNormalised);
         device->colorSpace()->normalisedChannelsValue(xItterator.rawData(), xNormalised);
 
         qreal z = 1.0;
         if (channelFlip[2]==true){
             z=-1.0;
         }
         QVector3D normal = QVector3D((xNormalised[channelToConvert]-0.5)*2, (yNormalised[channelToConvert]-0.5)*2, z);
         normal.normalize();
         finalNorm.fill(1.0);
         for (int c = 0; c<3; c++) {
             finalNorm[device->colorSpace()->channels().at(channelOrder[c])->displayPosition()] = (normal[channelOrder[c]]/2)+0.5;
         }
 
         finalNorm[alphaPos]= 1.0;
 
         quint8* pixel = finalIt.rawData();
         device->colorSpace()->fromNormalisedChannelsValue(pixel, finalNorm);
         memcpy(finalIt.rawData(), pixel, pixelSize);
 
     }
 }
diff --git a/plugins/filters/edgedetection/kis_edge_detection_filter.cpp b/plugins/filters/edgedetection/kis_edge_detection_filter.cpp
index 962bf437fb..4e8e8829b8 100644
--- a/plugins/filters/edgedetection/kis_edge_detection_filter.cpp
+++ b/plugins/filters/edgedetection/kis_edge_detection_filter.cpp
@@ -1,170 +1,177 @@
 /*
  * Copyright (c) 2017 Wolthera van Hövell tot Westerflier <griffinvalley@gmail.com>
  *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
  *  the Free Software Foundation; either version 2 of the License, or
  *  (at your option) any later version.
  *
  *  This program is distributed in the hope that it will be useful,
  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  *  GNU General Public License for more details.
  *
  *  You should have received a copy of the GNU General Public License
  *  along with this program; if not, write to the Free Software
  *  Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  */
 #include "kis_edge_detection_filter.h"
 #include "kis_wdg_edge_detection.h"
 #include <kis_edge_detection_kernel.h>
 #include <kis_convolution_kernel.h>
 #include <kis_convolution_painter.h>
 
 #include <KoColorSpaceRegistry.h>
 #include <KoColorModelStandardIds.h>
 
 #include <filter/kis_filter_category_ids.h>
 #include <filter/kis_filter_configuration.h>
 #include <kis_selection.h>
 #include <kis_paint_device.h>
 #include <kis_processing_information.h>
 #include "kis_lod_transform.h"
 
 #include <kpluginfactory.h>
 
 #include <klocalizedstring.h>
 #include <filter/kis_filter_registry.h>
 
 K_PLUGIN_FACTORY_WITH_JSON(KritaEdgeDetectionFilterFactory, "kritaedgedetection.json", registerPlugin<KritaEdgeDetectionFilter>();)
 
 KritaEdgeDetectionFilter::KritaEdgeDetectionFilter(QObject *parent, const QVariantList &)
     : QObject(parent)
 {
     KisFilterRegistry::instance()->add(KisFilterSP(new KisEdgeDetectionFilter()));
 }
 
 KritaEdgeDetectionFilter::~KritaEdgeDetectionFilter()
 {
 }
 
 KisEdgeDetectionFilter::KisEdgeDetectionFilter(): KisFilter(id(), FiltersCategoryEdgeDetectionId, i18n("&Edge Detection..."))
 {
     setSupportsPainting(true);
     setSupportsAdjustmentLayers(true);
     setSupportsLevelOfDetail(true);
     setColorSpaceIndependence(FULLY_INDEPENDENT);
     setShowConfigurationWidget(true);
 }
 
 void KisEdgeDetectionFilter::processImpl(KisPaintDeviceSP device, const QRect &rect, const KisFilterConfigurationSP config, KoUpdater *progressUpdater) const
 {
     Q_ASSERT(device != 0);
 
     KisFilterConfigurationSP configuration = config ? config : new KisFilterConfiguration(id().id(), 1);
 
     KisLodTransformScalar t(device);
 
     QVariant value;
     configuration->getProperty("horizRadius", value);
     float horizontalRadius = t.scale(value.toFloat());
     configuration->getProperty("vertRadius", value);
     float verticalRadius = t.scale(value.toFloat());
 
     QBitArray channelFlags;
     if (configuration) {
         channelFlags = configuration->channelFlags();
     }
 
     if (config->getString("type") == "canny") {
         KisFilterSP gaussian = KisFilterRegistry::instance()->get("gaussian blur");
         KisFilterConfigurationSP gaussianConf = gaussian->defaultConfiguration();
         gaussian->process(device, rect, gaussianConf, progressUpdater);
 
         int max = configuration->getInt("max_threshold", 200);
         int min = configuration->getInt("min_threshold", 100);
         KisEdgeDetectionKernel::applyCannyEdgeDetection(
             device, rect, horizontalRadius, verticalRadius, channelFlags, progressUpdater, max, min);
         return;
     }
 
     KisEdgeDetectionKernel::FilterType type = KisEdgeDetectionKernel::SobelVector;
     if (config->getString("type") == "prewitt") {
         type = KisEdgeDetectionKernel::Prewit;
     } else if (config->getString("type") == "simple") {
         type = KisEdgeDetectionKernel::Simple;
     }
 
     KisEdgeDetectionKernel::FilterOutput output = KisEdgeDetectionKernel::pythagorean;
     if (config->getString("output") == "xGrowth") {
         output = KisEdgeDetectionKernel::xGrowth;
     } else if (config->getString("output") == "xFall") {
         output = KisEdgeDetectionKernel::xFall;
     } else if (config->getString("output") == "yGrowth") {
         output = KisEdgeDetectionKernel::yGrowth;
     } else if (config->getString("output") == "yFall") {
         output = KisEdgeDetectionKernel::yFall;
     } else if (config->getString("output") == "radian") {
         output = KisEdgeDetectionKernel::radian;
     }
 
     KisEdgeDetectionKernel::applyEdgeDetection(device,
                                                rect,
                                                horizontalRadius,
                                                verticalRadius,
                                                type,
                                                channelFlags,
                                                progressUpdater,
                                                output,
                                                config->getBool("transparency", false));
 }
 
 KisFilterConfigurationSP KisEdgeDetectionFilter::defaultConfiguration() const
 {
     KisFilterConfigurationSP config = factoryConfiguration();
     config->setProperty("horizRadius", 1);
     config->setProperty("vertRadius", 1);
     config->setProperty("type", "prewitt");
     config->setProperty("output", "pythagorean");
     config->setProperty("lockAspect", true);
     config->setProperty("transparency", false);
     config->setProperty("max-threshold", 200);
     config->setProperty("min-threshold", 100);
 
     return config;
 }
 
 KisConfigWidget *KisEdgeDetectionFilter::createConfigurationWidget(QWidget *parent, const KisPaintDeviceSP dev, bool) const
 {
     Q_UNUSED(dev);
     return new KisWdgEdgeDetection(parent);
 }
 
 QRect KisEdgeDetectionFilter::neededRect(const QRect &rect, const KisFilterConfigurationSP _config, int lod) const
 {
     KisLodTransformScalar t(lod);
 
+    if(_config->getProperty("type") == "canny"){
+        return kisGrowRect(rect, 4);
+    }
     QVariant value;
     /**
      * NOTE: integer division by two is done on purpose,
      *       because the kernel size is always odd
      */
     const int halfWidth = _config->getProperty("horizRadius", value) ? KisEdgeDetectionKernel::kernelSizeFromRadius(t.scale(value.toFloat())) / 2 : 5;
     const int halfHeight = _config->getProperty("vertRadius", value) ? KisEdgeDetectionKernel::kernelSizeFromRadius(t.scale(value.toFloat())) / 2 : 5;
 
     return rect.adjusted(-halfWidth * 2, -halfHeight * 2, halfWidth * 2, halfHeight * 2);
 }
 
 QRect KisEdgeDetectionFilter::changedRect(const QRect &rect, const KisFilterConfigurationSP _config, int lod) const
 {
     KisLodTransformScalar t(lod);
 
     QVariant value;
 
+    if(_config->getProperty("type") == "canny"){
+        return kisGrowRect(rect, 4);
+    }
+
     const int halfWidth = _config->getProperty("horizRadius", value) ? KisEdgeDetectionKernel::kernelSizeFromRadius(t.scale(value.toFloat())) / 2 : 5;
     const int halfHeight = _config->getProperty("vertRadius", value) ? KisEdgeDetectionKernel::kernelSizeFromRadius(t.scale(value.toFloat())) / 2 : 5;
 
     return rect.adjusted( -halfWidth, -halfHeight, halfWidth, halfHeight);
 }
 
 #include "kis_edge_detection_filter.moc"