SUMMARY This was reported in the Krita Artists Forum: https://krita-artists.org/t/motion-blur-bug/1720 This happens for krita version 4.2.6 onwards to the latest Dec 08 Windows nightly portable .zip. Here is an examples file (100MB .kra) made with the 4.2.8 appimage: https://www.dropbox.com/s/wbfi3uton2q5byi/MBTilingOriginal.kra?dl=1 This contains two different 8-bit RGB images which have had the Motion Blur filter applied. Also, it has an edge detected (Prewitt, All Sides, 1x1) image to emphasise the artifacts. I used manual application rather that a filter layer in case the filter layer operations added any extra artifact effects. An edge detected version of the original image is included to ensure that the edge detection filter does not introduce artifacts, which it does not (but see Additional Notes a:). I tried it with the original images converted to 16-bit integer images and saw no noticable differences in the artifacts. (See Additional Notes b:) STEPS TO REPRODUCE 1. Apply a Motion Blur fliter to the image. If you adjust the Angle parameter (using the rotation dial at first) around the zero degrees value +/- 20 degrees you generally get line artifacts appearing. But they can appear all around the dial. 2. Fine adjustment of Angle value can make them worse or better. 3. Length values of 4 or less seem to give no artifacts. Length values of 5 or more do generally give artifacts but very large values of 30 or more do not, it seems. (Only limited testing was done.) 4. Use of the edge detection filter helps to emphasise and also reveal faint artifacts. OBSERVED RESULT See the attached .kra file EXPECTED RESULT There should be no artifacts SOFTWARE/OS VERSIONS Krita Version: 4.2.8 Languages: en_GB, en Hidpi: true Qt Version (compiled): 5.12.5 Version (loaded): 5.12.5 OS Information Build ABI: x86_64-little_endian-lp64 Build CPU: x86_64 CPU: x86_64 Kernel Type: linux Kernel Version: 4.19.0-6-amd64 Pretty Productname: Debian GNU/Linux 10 (buster) Product Type: debian Product Version: 10 OpenGL Info Vendor: "NVIDIA Corporation" Renderer: "GeForce GTX 750 Ti/PCIe/SSE2" Version: "4.6.0 NVIDIA 418.74" Shading language: "4.60 NVIDIA" Requested format: QSurfaceFormat(version 3.0, options QFlags<QSurfaceFormat::FormatOption>(DeprecatedFunctions), depthBufferSize 24, redBufferSize 8, greenBufferSize 8, blueBufferSize 8, alphaBufferSize 8, stencilBufferSize 8, samples -1, swapBehavior QSurfaceFormat::DoubleBuffer, swapInterval 0, colorSpace QSurfaceFormat::DefaultColorSpace, profile QSurfaceFormat::CompatibilityProfile) Current format: QSurfaceFormat(version 4.6, options QFlags<QSurfaceFormat::FormatOption>(DeprecatedFunctions), depthBufferSize 24, redBufferSize 8, greenBufferSize 8, blueBufferSize 8, alphaBufferSize 8, stencilBufferSize 8, samples -1, swapBehavior QSurfaceFormat::DoubleBuffer, swapInterval 0, colorSpace QSurfaceFormat::DefaultColorSpace, profile QSurfaceFormat::CompatibilityProfile) Version: 4.6 Supports deprecated functions true is OpenGL ES: false QPA OpenGL Detection Info supportsDesktopGL: true supportsOpenGLES: true isQtPreferOpenGLES: false Hardware Information GPU Acceleration: auto Memory: 16039 Mb Number of Cores: 8 Swap Location: /tmp Current Settings Current Swap Location: /tmp Undo Enabled: 1 Undo Stack Limit: 18 Use OpenGL: 1 Use OpenGL Texture Buffer: 1 Use AMD Vectorization Workaround: 0 Canvas State: OPENGL_SUCCESS Autosave Interval: 360 Use Backup Files: 1 Number of Backups Kept: 1 Backup File Suffix: ~ Backup Location: Same Folder as the File Use Win8 Pointer Input: 0 Use RightMiddleTabletButton Workaround: 0 Levels of Detail Enabled: 0 Use Zip64: 0 ADDITIONAL INFORMATION a: As used, with default values, the edge detection filter introduces no artifacts. If you use Formula:Any, Output:some Edge option, Radius:Any then you get horizontal artifacts at a 512 spacing, even on a plain colour image. This may need to be the subject of a separate bug report. b: With 16 bit images, the edge detection applied to the original images is notably different from the result of applying it to 8-bit images. This is not related to the purpose of this bug report and there may be a logical explanation but it seems strange.
From my testing is seem to happen from 4.2.0-preview onwards.
Git commit a5c1f8a0bf2d2d3ee7f4b1ad86ada04e6bc50e0c by Dmitry Kazakov. Committed on 10/03/2020 at 13:23. Pushed by dkazakov into branch 'master'. Fix artifacts in Motion Blur filter 1) Kernel size should be odd to do convolution right 2) halfWidth/halfHeight should be calculated consistently in all the three functions. Before the ptch there was a misprint in needRect and changeRect: both used 'cos' for both width and height. M +1 -1 libs/image/kis_lod_transform.h M +43 -53 plugins/filters/blur/kis_motion_blur_filter.cpp https://invent.kde.org/kde/krita/commit/a5c1f8a0bf2d2d3ee7f4b1ad86ada04e6bc50e0c