LogalDeveloper/Minecraft-Overviewer
Archived
0
Fork 0
Code Activity

added overall_alpha parameter to alpha_over, speeding up lighting

Browse Source
This commit is contained in:
Aaron Griffith
2011-03-26 21:52:16 -04:00
parent 0fe9c7c050
commit 1f27ccb9be
3 changed files with 39 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

83
src/composite.c
View File

@@ -58,52 +58,20 @@ imaging_python_to_c(PyObject *obj)
return image;
}
/**
* img should be an Image object, type 'L' or 'RGBA'
* for RGBA images, operates on the alpha only
* factor should be between 0 and 1, inclusive
*/
PyObject *brightness(PyObject *img, float factor) {
Imaging imDest;
int offset, stride, x, y, xsize, ysize;
imDest = imaging_python_to_c(img);
if (!imDest)
return NULL;
if (strcmp(imDest->mode, "RGBA") != 0 && strcmp(imDest->mode, "L") != 0) {
PyErr_SetString(PyExc_ValueError,
"given image does not have mode \"RGBA\" or \"L\"");
return NULL;
}
/* how far into image the first alpha byte resides */
offset = (imDest->pixelsize == 4 ? 3 : 0);
/* how many bytes to skip to get to the next alpha byte */
stride = imDest->pixelsize;
xsize = imDest->xsize;
ysize = imDest->ysize;
for (y = 0; y < ysize; y++) {
UINT8 *out = (UINT8 *)imDest->image[y] + offset;
for (x = 0; x < xsize; x++) {
*out *= factor;
out += stride;
}
}
return NULL;
/* convenience alpha_over with 1.0 as overall_alpha */
inline PyObject* alpha_over(PyObject *dest, PyObject *src, PyObject *mask,
int dx, int dy, int xsize, int ysize) {
return alpha_over_full(dest, src, mask, 1.0f, dx, dy, xsize, ysize);
}
/* the alpha_over function, in a form that can be called from C */
/* if xsize, ysize are negative, they are instead set to the size of the image in src */
/* returns NULL on error, dest on success. You do NOT need to decref the return! */
PyObject *
alpha_over(PyObject *dest, PyObject *src, PyObject *mask, int dx, int dy,
int xsize, int ysize)
{
/* the full alpha_over function, in a form that can be called from C
* overall_alpha is multiplied with the whole mask, useful for lighting...
* if xsize, ysize are negative, they are instead set to the size of the image in src
* returns NULL on error, dest on success. You do NOT need to decref the return!
*/
inline PyObject *
alpha_over_full(PyObject *dest, PyObject *src, PyObject *mask, float overall_alpha,
int dx, int dy, int xsize, int ysize) {
/* libImaging handles */
Imaging imDest, imSrc, imMask;
/* cached blend properties */
@@ -114,6 +82,12 @@ alpha_over(PyObject *dest, PyObject *src, PyObject *mask, int dx, int dy,
unsigned int x, y, i;
/* temporary calculation variables */
int tmp1, tmp2, tmp3;
/* integer [0, 255] version of overall_alpha */
UINT8 overall_alpha_int = 255 * overall_alpha;
/* short-circuit this whole thing if overall_alpha is zero */
if (overall_alpha_int == 0)
return dest;
imDest = imaging_python_to_c(dest);
imSrc = imaging_python_to_c(src);
@@ -202,9 +176,18 @@ alpha_over(PyObject *dest, PyObject *src, PyObject *mask, int dx, int dy,
UINT8 *inmask = (UINT8 *)imMask->image[sy + y] + sx * mask_stride + mask_offset;
for (x = 0; x < xsize; x++) {
UINT8 in_alpha;
/* apply overall_alpha */
if (overall_alpha_int != 255 && *inmask != 0) {
in_alpha = MULDIV255(*inmask, overall_alpha_int, tmp1);
} else {
in_alpha = *inmask;
}
/* special cases */
if (*inmask == 255 || *outmask == 0) {
*outmask = *inmask;
if (in_alpha == 255 || *outmask == 0) {
*outmask = in_alpha;
*out = *in;
out++, in++;
@@ -213,18 +196,18 @@ alpha_over(PyObject *dest, PyObject *src, PyObject *mask, int dx, int dy,
*out = *in;
out++, in++;
}
else if (*inmask == 0) {
else if (in_alpha == 0) {
/* do nothing -- source is fully transparent */
out += 3;
in += 3;
}
else {
/* general case */
int alpha = *inmask + MULDIV255(*outmask, 255 - *inmask, tmp1);
int alpha = in_alpha + MULDIV255(*outmask, 255 - in_alpha, tmp1);
for (i = 0; i < 3; i++) {
/* general case */
*out = MULDIV255(*in, *inmask, tmp1) +
MULDIV255(MULDIV255(*out, *outmask, tmp2), 255 - *inmask, tmp3);
*out = MULDIV255(*in, in_alpha, tmp1) +
MULDIV255(MULDIV255(*out, *outmask, tmp2), 255 - in_alpha, tmp3);
*out = (*out * 255) / alpha;
out++, in++;