This repository has been archived on 2023-10-09. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
blender-archive/source/blender/compositor/intern/COM_MemoryBuffer.h
T
Lukas Tönne 67134a7bf6 Fix for EWA (elliptical weighted average) sampling in the compositor.
EWA sampling is designed for downsampling images, i.e. scaling down the size of
input image pixels, which happens regularly in compositing. While the standard
sampling methods (linear, cubic) work reasonably well for linear
transformations, they don't yield good results in non-linear cases like
perspective projection or arbitrary displacement. EWA sampling is comparable to
mipmapping, but avoids problems with discontinuities.

To work correctly the EWA algorithm needs partial derivatives of the mapping
functions which convert output pixel coordinates back into the input image
space (2x2 Jacobian matrix). With these derivatives the EWA algorithm
projects ellipses into the input space and accumulates colors over their
area. This calculation was not done correctly in the compositor, only the
derivatives du/dx and dv/dy were calculation, basically this means it only
worked for non-rotated input images.

The patch introduces full derivative calculations du/dx, du/dy, dv/dx, dv/dy for
the 3 nodes which use EWA sampling currently: PlaneTrackWarp, MapUV and
Displace. In addition the calculation of ellipsis area and axis-aligned
bounding boxes has been fixed.

For the MapUV and Displace nodes the derivatives have to be estimated by
evaluating the UV/displacement inputs with 1-pixel offsets, which can still have
problems on discontinuities and sub-pixel variations. These potential problems
can only be alleviated by more radical design changes in the compositor
functions, which are out of scope for now. Basically the values passed to the
UV/Displacement inputs would need to be associated with their 1st order
derivatives, which requires a general approach to derivatives in all nodes.
2013-12-04 16:05:56 +01:00

298 lines
7.7 KiB
C++

/*
* Copyright 2011, Blender Foundation.
*
* 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.
*
* Contributor:
* Jeroen Bakker
* Monique Dewanchand
*/
class MemoryBuffer;
#ifndef _COM_MemoryBuffer_h_
#define _COM_MemoryBuffer_h_
#include "COM_ExecutionGroup.h"
#include "COM_MemoryProxy.h"
extern "C" {
#include "BLI_math.h"
#include "BLI_rect.h"
}
/**
* @brief state of a memory buffer
* @ingroup Memory
*/
typedef enum MemoryBufferState {
/** @brief memory has been allocated on creator device and CPU machine, but kernel has not been executed */
COM_MB_ALLOCATED = 1,
/** @brief memory is available for use, content has been created */
COM_MB_AVAILABLE = 2,
/** @brief chunk is consolidated from other chunks. special state.*/
COM_MB_TEMPORARILY = 6
} MemoryBufferState;
typedef enum MemoryBufferExtend {
COM_MB_CLIP,
COM_MB_EXTEND,
COM_MB_REPEAT
} MemoryBufferExtend;
class MemoryProxy;
/**
* @brief a MemoryBuffer contains access to the data of a chunk
*/
class MemoryBuffer {
private:
/**
* @brief proxy of the memory (same for all chunks in the same buffer)
*/
MemoryProxy *m_memoryProxy;
/**
* @brief the type of buffer COM_DT_VALUE, COM_DT_VECTOR, COM_DT_COLOR
*/
DataType m_datatype;
/**
* @brief region of this buffer inside relative to the MemoryProxy
*/
rcti m_rect;
/**
* brief refers to the chunknumber within the executiongroup where related to the MemoryProxy
* @see memoryProxy
*/
unsigned int m_chunkNumber;
/**
* @brief width of the chunk
*/
unsigned int m_chunkWidth;
/**
* @brief state of the buffer
*/
MemoryBufferState m_state;
/**
* @brief the actual float buffer/data
*/
float *m_buffer;
public:
/**
* @brief construct new MemoryBuffer for a chunk
*/
MemoryBuffer(MemoryProxy *memoryProxy, unsigned int chunkNumber, rcti *rect);
/**
* @brief construct new temporarily MemoryBuffer for an area
*/
MemoryBuffer(MemoryProxy *memoryProxy, rcti *rect);
/**
* @brief destructor
*/
~MemoryBuffer();
/**
* @brief read the ChunkNumber of this MemoryBuffer
*/
unsigned int getChunkNumber() { return this->m_chunkNumber; }
/**
* @brief get the data of this MemoryBuffer
* @note buffer should already be available in memory
*/
float *getBuffer() { return this->m_buffer; }
/**
* @brief after execution the state will be set to available by calling this method
*/
void setCreatedState()
{
this->m_state = COM_MB_AVAILABLE;
}
inline void wrap_pixel(int &x, int &y, MemoryBufferExtend extend_x, MemoryBufferExtend extend_y)
{
int w = m_rect.xmax - m_rect.xmin;
int h = m_rect.ymax - m_rect.ymin;
x = x - m_rect.xmin;
y = y - m_rect.ymin;
switch (extend_x) {
case COM_MB_CLIP:
break;
case COM_MB_EXTEND:
if (x < 0) x = 0;
if (x >= w) x = w;
break;
case COM_MB_REPEAT:
x = (x >= 0.0f ? (x % w) : (x % w) + w);
break;
}
switch (extend_y) {
case COM_MB_CLIP:
break;
case COM_MB_EXTEND:
if (y < 0) y = 0;
if (y >= h) y = h;
break;
case COM_MB_REPEAT:
y = (y >= 0.0f ? (y % h) : (y % h) + h);
break;
}
}
inline void read(float result[4], int x, int y,
MemoryBufferExtend extend_x = COM_MB_CLIP,
MemoryBufferExtend extend_y = COM_MB_CLIP)
{
bool clip_x = (extend_x == COM_MB_CLIP && (x < m_rect.xmin || x >= m_rect.xmax));
bool clip_y = (extend_y == COM_MB_CLIP && (y < m_rect.ymin || y >= m_rect.ymax));
if (clip_x || clip_y) {
/* clip result outside rect is zero */
zero_v4(result);
}
else {
wrap_pixel(x, y, extend_x, extend_y);
const int offset = (this->m_chunkWidth * y + x) * COM_NUMBER_OF_CHANNELS;
copy_v4_v4(result, &this->m_buffer[offset]);
}
}
inline void readNoCheck(float result[4], int x, int y,
MemoryBufferExtend extend_x = COM_MB_CLIP,
MemoryBufferExtend extend_y = COM_MB_CLIP)
{
wrap_pixel(x, y, extend_x, extend_y);
const int offset = (this->m_chunkWidth * y + x) * COM_NUMBER_OF_CHANNELS;
BLI_assert(offset >= 0);
BLI_assert(offset < this->determineBufferSize() * COM_NUMBER_OF_CHANNELS);
BLI_assert(!(extend_x == COM_MB_CLIP && (x < m_rect.xmin || x >= m_rect.xmax)) &&
!(extend_y == COM_MB_CLIP && (y < m_rect.ymin || y >= m_rect.ymax)));
#if 0
/* always true */
BLI_assert((int)(MEM_allocN_len(this->m_buffer) / sizeof(*this->m_buffer)) ==
(int)(this->determineBufferSize() * COM_NUMBER_OF_CHANNELS));
#endif
copy_v4_v4(result, &this->m_buffer[offset]);
}
void writePixel(int x, int y, const float color[4]);
void addPixel(int x, int y, const float color[4]);
inline void readBilinear(float result[4], float x, float y,
MemoryBufferExtend extend_x = COM_MB_CLIP,
MemoryBufferExtend extend_y = COM_MB_CLIP)
{
int x1 = floor(x);
int y1 = floor(y);
int x2 = x1 + 1;
int y2 = y1 + 1;
wrap_pixel(x1, y1, extend_x, extend_y);
wrap_pixel(x2, y2, extend_x, extend_y);
float valuex = x - x1;
float valuey = y - y1;
float mvaluex = 1.0f - valuex;
float mvaluey = 1.0f - valuey;
float color1[4];
float color2[4];
float color3[4];
float color4[4];
read(color1, x1, y1);
read(color2, x1, y2);
read(color3, x2, y1);
read(color4, x2, y2);
color1[0] = color1[0] * mvaluey + color2[0] * valuey;
color1[1] = color1[1] * mvaluey + color2[1] * valuey;
color1[2] = color1[2] * mvaluey + color2[2] * valuey;
color1[3] = color1[3] * mvaluey + color2[3] * valuey;
color3[0] = color3[0] * mvaluey + color4[0] * valuey;
color3[1] = color3[1] * mvaluey + color4[1] * valuey;
color3[2] = color3[2] * mvaluey + color4[2] * valuey;
color3[3] = color3[3] * mvaluey + color4[3] * valuey;
result[0] = color1[0] * mvaluex + color3[0] * valuex;
result[1] = color1[1] * mvaluex + color3[1] * valuex;
result[2] = color1[2] * mvaluex + color3[2] * valuex;
result[3] = color1[3] * mvaluex + color3[3] * valuex;
}
void readEWA(float result[4], const float uv[2], const float derivatives[2][2], PixelSampler sampler);
/**
* @brief is this MemoryBuffer a temporarily buffer (based on an area, not on a chunk)
*/
inline const bool isTemporarily() const { return this->m_state == COM_MB_TEMPORARILY; }
/**
* @brief add the content from otherBuffer to this MemoryBuffer
* @param otherBuffer source buffer
*
* @note take care when running this on a new buffer since it wont fill in
* uninitialized values in areas where the buffers don't overlap.
*/
void copyContentFrom(MemoryBuffer *otherBuffer);
/**
* @brief get the rect of this MemoryBuffer
*/
rcti *getRect() { return &this->m_rect; }
/**
* @brief get the width of this MemoryBuffer
*/
int getWidth() const;
/**
* @brief get the height of this MemoryBuffer
*/
int getHeight() const;
/**
* @brief clear the buffer. Make all pixels black transparent.
*/
void clear();
MemoryBuffer *duplicate();
float *convertToValueBuffer();
float getMaximumValue();
float getMaximumValue(rcti *rect);
private:
unsigned int determineBufferSize();
#ifdef WITH_CXX_GUARDEDALLOC
MEM_CXX_CLASS_ALLOC_FUNCS("COM:MemoryBuffer")
#endif
};
#endif