resolved a crash when using integrated deep front and back separately.
changed how inputs were requested to ensure correct handling of passed through layers/channels.
updated to support C++11 for Nuke 11.3 build.
refactored how input channels were handled to accommodate instances where “all” was requested. This will ultimately fix issues related to a higher number of channels being convolved due to how channel indicies are created.
added C++11 support for Nuke 11.0.
updated build settings for Nuke 11.0 compatibility.
re-factored caching calls within engine to avoid access deadlock
added support for a PG_BOKEH_ABORT_ON_LICENSE_FAIL environment variable that, if set to 1, will cause Bokeh to abort if a license isn’t available.
fixed a bug introduced in 1.4.0 which caused Bokeh to render at sub-optimal speed.
updated log output to print to stdout.
Bokeh will now watermark the output when no license is valid vs. aborting the render.
updates to support Nuke 9.0
changes to kernel generation and initialization to avoid locking issues.
fixed a bug that was causing way to many threads to be created.
Deep Front is used by default for deep image depths, Deep Back may be used if selected from the z channel parameter.
Bokeh now uses Peregrine’s deep combine to avoid data clipping introduced in Nuke’s API.
added Integrate Deep Front and Back Seperately parameter to toggle 1.3.5’s update on/off
default corrective slices set to 10
added missing kernel channel to specify a fallback channel for an input kernel when it is missing a required channel
integrated peregrine deep combining engine to support correct volume defocusing
updated to latest Nuke 7.0 build
fixed a bug that would lead to some value combinations causing a black image output
fixed volumentric compositing of deep defocus samples
stereo exr writes were failing when using a deep image input ( it was writing the same eye twice )
fixed bug which caused no output when input channels were set to All with deep input
initial Nuke 7.0 support
Y stretching when data window covers negative values
refactored thread locking
fixed bug with incorrectly initialized chromic abberation toggle
Bokeh wasn’t correctly computing the max kernel size which was causing slower rendering, this has been corrected
added the option of supplying a mattes input which can contain matte channels for controlling Bokeh
added Bloom Matte Channel to control bloom levels via an input matte
added Defocus Matte Channel to control defocus levels via an input matte
updated chromic aberration to be a product of the lens vs. the kernel
added Chromic Aberration Matte Channel to control CA level via an input matte
added Optical Artifacts tab to re-organize UI
moved normalize visualization to the Bokeh tab
updated documentation to reflect UI changes and new features
fixed correctie slices compositing artifacts
resolved caching issues which caused slow updates on initial read
front & back multiplier now supported when real world lens mode is selected
changed Bokeh Channels to Input_ChannelSet_knob to give access to an “all” set
refactored corrective slices and promoted them to their own tabs
added world scale multiplier to accomodate for working scales such as 10cm = 1unit
replaced polygonal kernel style with Aperture Blades which more closely mimics the blades within a lens
added blade curvature to kernel parameters
more accurate proxy handling
fixed cache slow downs
fixed issues causing slow initial renders when roto and scanline renderer connected
improved corrective slice compositing
fixed issue where deep images we’re being forced to validate by nuke, so we have to handle that internally
tried to speed up pre-pass calls with an input ScanlineRender
fixed issue where front/back region output was incorrect
changed how nan values are handled
fixed bug where image would render black if focal plane was equal to the only depth value in the image
fixed deep image issues where out of range pixels were sorting incorrectly leading to vector::_M_fill_insert errors
added “expand to bounding box” option in the kernel tab to adjust the images output bbox to fit the max kernel size
fixed bug where inf depth values would result in erroneous depth slice results
fixed alpha compositing with deep images that resolves subtle alpha artifacts
improved corrective slices for deep images
bloom works again
fixed corrective slices ( to more closely match 1.2.1 )
resolved cache issues with deep images ( causing “can’t access deep row” errors )
fixed update crash, this caused instability when adjusting parameters while Bokeh was still updating
fixed bug that caused Bokeh to crash when the input was disconnected while updating
refactored how the plugin handles multiple threads to avoid locking
cleaned up interface to avoid screen space abuse
added custom filmback size and disabled aperture w & h when not selected
added feet and inches as a world scale option
refactored real world lens equations
removed lens style option in favour of a more correct default
we’ve removed the zeiss formula input and reverted back to the traditional ( 1730 ) value, most of our newly derived equations are based on this. If you need it back for some reason email firstname.lastname@example.org.
solved minute kernel coefficient mismatching artifacts
rewrote kernel scaling routines ( defaults to true cubic )
deep images would crash Bokeh if there were no samples in a row
“zero” depth rows correctly handled
input kernel normalization based on max summation resuting in more correct reproduction of input kernels with varying colour intensities
focus distance visualization normalized when using real world lens simulation
all knobs which are invalid when a camera is connected are now disabled
added kernel aspect ratio option, this allows asymetric kernels for matching anamorphic bokeh among other things
fixed issue with bokeh sometimes complaining it can’t access the deep image row
fixed input kernel issues
fixed issue with black output when depth of the whole image == focal plane
fixed the fact film format wasn’t working when a 3d camera was used
spherical abberation now applied to polygonal kernels correctly
new documentation ( what you’re reading )
moved changelog into documentation
added depth style parameter, used to control whether to interperate depth as real depth values or 1/z values ( as produced by Nukes scanline renderer )
added option to simulate lens that will match PRMan’s defocus
supports all channels in deep data ( vs. just alpha )
changed default polygonal kernel side count to 6
fixed offset issues with chromic abberation
chromic abberation handled more subtley
improced lens simulation accuracy
added parameters to control the front and back defocus levels seperately
weight buffer incorrectly normalizing output values when using deep image input
fixed issue with kernel generation causing “swimming” in animated focus pulls
fixed deep image depth reading which returned inf in some instances
deep image support would crash if not fed through a DeepToImage, this is fixed now
deep image support
update build system to handle multiple nuke versions
correctly accounted for proxy scale in non-bokeh mode
crash when depth slices is set to 0
resolved an issue where rendering would over saturation requests to the nuke cache causing much slower renders
removed setting knobs based on input cameras to avoid polluting the hash tree
removed limit on depth slices
added aperture width/heigh for real world lens simulation
input cameras height and width now sets aperture height and width on bokeh
renamed film back to film format
replaced still camera with motion picture formats
more accurate lens simulation
added a parameter for the desired Zeiss formula
cameras fstop now passed to Bokeh correctly
added camera input to control a limited set of parameters
multiplier is ignored when real world lens is selected
added the option to output just the front or back focus planes, when added they produce the same result as the full defocus effect
added focus matte output mode
added world scale parameter to control how real world lens interperates depth values
added film back parameter to real world lens
fixed varying channel issues due to requesting invalid channels
crushed real world lens bug
colourized depth preview
changed range for defocus level
sped up kernel creation
optimized depth and kernel preview
add the option of specifying what channel to use for depth
fixed crash when viewing channels other than RGBA
fixed polygon sides issue ( was just drawing a square )
reorganized CMake and directory structure
first build on Windows, wehay!
sped up preview depth values mode