Additional functions have been put in to support planar images. They are the following:
VideoFrame:
int GetPitch(int plane)
int GetRowSize(int plane)
int GetHeight(int plane)
VideoFrame* Subframe(int rel_offset, int new_pitch, int new_row_size, int new_height, int !rel_offsetU, int !rel_offsetV, int !pitchUV)
const BYTE* GetReadPtr(int plane)
BYTE* GetWritePtr(int plane)
Where plane can be:
- PLANAR_Y
- PLANAR_U
- PLANAR_V
PLANAR_Y is equivalent to calling the old functions.
It can be determined whether or not you are dealing with a planar images by using vi.IsPlanar(). This will return true, if image is planar. Using vi.RowSize()from VideoInfo only provides you with correct numbers for the first plane. Use RowSize(int plane) from VideoFrame to obtain the proper results. The same goes for vi.BytesFromPixels(), which also only returns correct results for the first plane.
What can be safely assumed for planar images is:
- * Pitch U = Pitch V.
- * RowSize U = RowSize V
- * Height U = Height V
You should consider the planes as seperate memory regions, and you shouldn't begin guessing where "U would be relative to Y", since it might as well change.
The general guideline should (as always) be: Use the functions - never directly access the members of VideoInfo or VideoFrameBuffer or VideoFrame structs directly, unless you are creating an input source. Always use the supplied funtions! This is the only way for us to be able to maintain compatibility with future releases. All processing filters should be able to use the methods exclusively.
GetRowSize() is special, since it can return mod16 widths for Luma and mod8 aligned widths for chroma planes. When you request the rowsize, you can add the parameter "PLANAR_Y_ALIGNED" or "PLANAR_U_ALIGNED"/"PLANAR_V_ALIGNED". This will give you the width of the image, rounded up to nearest mod16 (or if already mod16, just rowsize as usual).
This means that you can request an image larger than the actual image, without risking to write or read into the next line. You can safely do whatever you like with this data, since AviSynth completely ignores it. This will enable you to operate on 8 bytes (1 quadword) at the time for all planes.
Some (producing) filters may want to set a forced pitch. Per default, when you request a videoframe with a specific modX pitch it might be overruled by AviSynth.(You do this by using env->NewVideoFrame(VideoInfo, [alignment])) In this (very rare case), you can use a negative alignment. This will make the alignment forced.
If you force an alignment you MUST also apply the AlignPlanar(!PClip) filter after your own, otherwise your filter will not produce valid output. See source.cpp for examples.
See:
The example of an AviSynyh TwoFiveInvert filter.
Something about the specific ColorSpaces.
More about DataStorageInAviSynth.
A general introduction to WorkingWithImages.
