diff options
Diffstat (limited to 'source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp')
-rw-r--r-- | source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp | 134 |
1 files changed, 67 insertions, 67 deletions
diff --git a/source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp b/source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp index c3c115b7c3b..e5a145bab2e 100644 --- a/source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp +++ b/source/blender/compositor/operations/COM_DoubleEdgeMaskOperation.cpp @@ -781,10 +781,10 @@ static void do_allEdgeDetection(unsigned int t, unsigned int rw, unsigned int *l if (!limask[a]) { // if the inner mask is empty if (lomask[a]) { // if the outer mask is full /* - Next we test all 4 directions around the current pixel: next/prev/up/down - The test ensures that the outer mask is empty and that the inner mask - is also empty. If both conditions are true for any one of the 4 adjacent pixels - then the current pixel is counted as being a true outer edge pixel. + * Next we test all 4 directions around the current pixel: next/prev/up/down + * The test ensures that the outer mask is empty and that the inner mask + * is also empty. If both conditions are true for any one of the 4 adjacent pixels + * then the current pixel is counted as being a true outer edge pixel. */ if ((!lomask[pix_nextCol] && !limask[pix_nextCol]) || (!lomask[pix_prevCol] && !limask[pix_prevCol]) || @@ -843,10 +843,10 @@ static void do_adjacentEdgeDetection(unsigned int t, unsigned int rw, unsigned i if (!limask[a]) { // if the inner mask is empty if (lomask[a]) { // if the outer mask is full /* - Next we test all 4 directions around the current pixel: next/prev/up/down - The test ensures that the outer mask is empty and that the inner mask - is also empty. If both conditions are true for any one of the 4 adjacent pixels - then the current pixel is counted as being a true outer edge pixel. + * Next we test all 4 directions around the current pixel: next/prev/up/down + * The test ensures that the outer mask is empty and that the inner mask + * is also empty. If both conditions are true for any one of the 4 adjacent pixels + * then the current pixel is counted as being a true outer edge pixel. */ if ((!lomask[pix_nextCol] && !limask[pix_nextCol]) || (!lomask[pix_prevCol] && !limask[pix_prevCol]) || @@ -902,65 +902,65 @@ static void do_createEdgeLocationBuffer(unsigned int t, unsigned int rw, unsigne unsigned int outerAccum = 0; // for looping outer edge pixel indexes, represents current position from offset unsigned int gradientAccum = 0; // for looping gradient pixel indexes, represents current position from offset /* - Here we compute the size of buffer needed to hold (row,col) coordinates - for each pixel previously determined to be either gradient, inner edge, - or outer edge. - - Allocation is done by requesting 4 bytes "sizeof(int)" per pixel, even - though gbuf[] is declared as unsigned short* (2 bytes) because we don't - store the pixel indexes, we only store x,y location of pixel in buffer. - - This does make the assumption that x and y can fit in 16 unsigned bits - so if Blender starts doing renders greater than 65536 in either direction - this will need to allocate gbuf[] as unsigned int *and allocate 8 bytes - per flagged pixel. - - In general, the buffer on-screen: - - Example: 9 by 9 pixel block - - . = pixel non-white in both outer and inner mask - o = pixel white in outer, but not inner mask, adjacent to "." pixel - g = pixel white in outer, but not inner mask, not adjacent to "." pixel - i = pixel white in inner mask, adjacent to "g" or "." pixel - F = pixel white in inner mask, only adjacent to other pixels white in the inner mask - - - ......... <----- pixel #80 - ..oooo... - .oggggo.. - .oggiggo. - .ogiFigo. - .oggiggo. - .oggggo.. - ..oooo... - pixel #00 -----> ......... - - gsz = 18 (18 "g" pixels above) - isz = 4 (4 "i" pixels above) - osz = 18 (18 "o" pixels above) - - - The memory in gbuf[] after filling will look like this: - - gradientFillOffset (0 pixels) innerEdgeOffset (18 pixels) outerEdgeOffset (22 pixels) - / / / - / / / - |X Y X Y X Y X Y > <X Y X Y > <X Y X Y X Y > <X Y X Y | <- (x,y) - +--------------------------------> <----------------> <------------------------> <----------------+ - |0 2 4 6 8 10 12 14 > ... <68 70 72 74 > ... <80 82 84 86 88 90 > ... <152 154 156 158 | <- bytes - +--------------------------------> <----------------> <------------------------> <----------------+ - |g0 g0 g1 g1 g2 g2 g3 g3 > <g17 g17 i0 i0 > <i2 i2 i3 i3 o0 o0 > <o16 o16 o17 o17 | <- pixel - / / / - / / / - / / / - +---------- gradientAccum (18) ---------+ +--- innerAccum (22) ---+ +--- outerAccum (40) ---+ - - - Ultimately we do need the pixel's memory buffer index to set the output - pixel color, but it's faster to reconstruct the memory buffer location - each iteration of the final gradient calculation than it is to deconstruct - a memory location into x,y pairs each round. + * Here we compute the size of buffer needed to hold (row,col) coordinates + * for each pixel previously determined to be either gradient, inner edge, + * or outer edge. + * + * Allocation is done by requesting 4 bytes "sizeof(int)" per pixel, even + * though gbuf[] is declared as unsigned short* (2 bytes) because we don't + * store the pixel indexes, we only store x,y location of pixel in buffer. + * + * This does make the assumption that x and y can fit in 16 unsigned bits + * so if Blender starts doing renders greater than 65536 in either direction + * this will need to allocate gbuf[] as unsigned int *and allocate 8 bytes + * per flagged pixel. + * + * In general, the buffer on-screen: + * + * Example: 9 by 9 pixel block + * + * . = pixel non-white in both outer and inner mask + * o = pixel white in outer, but not inner mask, adjacent to "." pixel + * g = pixel white in outer, but not inner mask, not adjacent to "." pixel + * i = pixel white in inner mask, adjacent to "g" or "." pixel + * F = pixel white in inner mask, only adjacent to other pixels white in the inner mask + * + * + * ......... <----- pixel #80 + * ..oooo... + * .oggggo.. + * .oggiggo. + * .ogiFigo. + * .oggiggo. + * .oggggo.. + * ..oooo... + * pixel #00 -----> ......... + * + * gsz = 18 (18 "g" pixels above) + * isz = 4 (4 "i" pixels above) + * osz = 18 (18 "o" pixels above) + * + * + * The memory in gbuf[] after filling will look like this: + * + * gradientFillOffset (0 pixels) innerEdgeOffset (18 pixels) outerEdgeOffset (22 pixels) + * / / / + * / / / + * |X Y X Y X Y X Y > <X Y X Y > <X Y X Y X Y > <X Y X Y | <- (x,y) + * +--------------------------------> <----------------> <------------------------> <----------------+ + * |0 2 4 6 8 10 12 14 > ... <68 70 72 74 > ... <80 82 84 86 88 90 > ... <152 154 156 158 | <- bytes + * +--------------------------------> <----------------> <------------------------> <----------------+ + * |g0 g0 g1 g1 g2 g2 g3 g3 > <g17 g17 i0 i0 > <i2 i2 i3 i3 o0 o0 > <o16 o16 o17 o17 | <- pixel + * / / / + * / / / + * / / / + * +---------- gradientAccum (18) ---------+ +--- innerAccum (22) ---+ +--- outerAccum (40) ---+ + * + * + * Ultimately we do need the pixel's memory buffer index to set the output + * pixel color, but it's faster to reconstruct the memory buffer location + * each iteration of the final gradient calculation than it is to deconstruct + * a memory location into x,y pairs each round. */ |