static void streaks(NodeGlare* ndg, CompBuf* dst, CompBuf* src)
{
CompBuf *bsrc, *tsrc, *tdst, *sbuf;
int x, y, n;
unsigned int nump=0;
fRGB c1, c2, c3, c4;
float a, ang = 360.f/(float)ndg->angle;
bsrc = BTP(src, ndg->threshold, 1 << ndg->quality);
tsrc = dupalloc_compbuf(bsrc); // sample from buffer
tdst = alloc_compbuf(tsrc->x, tsrc->y, tsrc->type, 1); // sample to buffer
sbuf = alloc_compbuf(tsrc->x, tsrc->y, tsrc->type, 1); // streak sum buffer
for (a=0.f; a<360.f; a+=ang) {
const float an = (a + (float)ndg->angle_ofs)*(float)M_PI/180.f;
const float vx = cos((double)an), vy = sin((double)an);
for (n=0; n<ndg->iter; ++n) {
const float p4 = pow(4.0, (double)n);
const float vxp = vx*p4, vyp = vy*p4;
const float wt = pow((double)ndg->fade, (double)p4);
const float cmo = 1.f - pow((double)ndg->colmod, (double)n+1); // colormodulation amount relative to current pass
float* tdstcol = tdst->rect;
for (y=0; y<tsrc->y; ++y) {
for (x=0; x<tsrc->x; ++x, tdstcol+=4) {
// first pass no offset, always same for every pass, exact copy,
// otherwise results in uneven brightness, only need once
if (n==0) qd_getPixel(tsrc, x, y, c1); else c1[0]=c1[1]=c1[2]=0;
qd_getPixelLerp(tsrc, x + vxp, y + vyp, c2);
qd_getPixelLerp(tsrc, x + vxp*2.f, y + vyp*2.f, c3);
qd_getPixelLerp(tsrc, x + vxp*3.f, y + vyp*3.f, c4);
// modulate color to look vaguely similar to a color spectrum
fRGB_rgbmult(c2, 1.f, cmo, cmo);
fRGB_rgbmult(c3, cmo, cmo, 1.f);
fRGB_rgbmult(c4, cmo, 1.f, cmo);
tdstcol[0] = 0.5f*(tdstcol[0] + c1[0] + wt*(c2[0] + wt*(c3[0] + wt*c4[0])));
tdstcol[1] = 0.5f*(tdstcol[1] + c1[1] + wt*(c2[1] + wt*(c3[1] + wt*c4[1])));
tdstcol[2] = 0.5f*(tdstcol[2] + c1[2] + wt*(c2[2] + wt*(c3[2] + wt*c4[2])));
}
}
memcpy(tsrc->rect, tdst->rect, sizeof(float)*tdst->x*tdst->y*tdst->type);
}
addImage(sbuf, tsrc, 1.f/(float)(6 - ndg->iter));
memset(tdst->rect, 0, tdst->x*tdst->y*tdst->type*sizeof(float));
memcpy(tsrc->rect, bsrc->rect, bsrc->x*bsrc->y*bsrc->type*sizeof(float));
nump++;
}
mixImages(dst, sbuf, 0.5f + 0.5f*ndg->mix);
free_compbuf(tsrc);
free_compbuf(tdst);
free_compbuf(sbuf);
free_compbuf(bsrc);
}
// mix two images, src buffer does not have to be same size,
static void mixImages(CompBuf *dst, CompBuf *src, float mix)
{
int x, y;
fRGB c1, c2, *dcolp, *scolp;
const float mf = 2.f - 2.f*fabsf(mix - 0.5f);
if ((dst->x == src->x) && (dst->y == src->y)) {
for (y=0; y<dst->y; y++) {
dcolp = (fRGB*)&dst->rect[y*dst->x*dst->type];
scolp = (fRGB*)&src->rect[y*dst->x*dst->type];
for (x=0; x<dst->x; x++) {
fRGB_copy(c1, dcolp[x]);
fRGB_copy(c2, scolp[x]);
c1[0] += mix*(c2[0] - c1[0]);
c1[1] += mix*(c2[1] - c1[1]);
c1[2] += mix*(c2[2] - c1[2]);
if (c1[0] < 0.f) c1[0] = 0.f;
if (c1[1] < 0.f) c1[1] = 0.f;
if (c1[2] < 0.f) c1[2] = 0.f;
fRGB_mult(c1, mf);
fRGB_copy(dcolp[x], c1);
}
}
}
else {
float xr = src->x / (float)dst->x;
float yr = src->y / (float)dst->y;
for (y=0; y<dst->y; y++) {
dcolp = (fRGB*)&dst->rect[y*dst->x*dst->type];
for (x=0; x<dst->x; x++) {
fRGB_copy(c1, dcolp[x]);
qd_getPixelLerp(src, (x + 0.5f)*xr - 0.5f, (y + 0.5f)*yr - 0.5f, c2);
c1[0] += mix*(c2[0] - c1[0]);
c1[1] += mix*(c2[1] - c1[1]);
c1[2] += mix*(c2[2] - c1[2]);
if (c1[0] < 0.f) c1[0] = 0.f;
if (c1[1] < 0.f) c1[1] = 0.f;
if (c1[2] < 0.f) c1[2] = 0.f;
fRGB_mult(c1, mf);
fRGB_copy(dcolp[x], c1);
}
}
}
}
static void ghosts(NodeGlare* ndg, CompBuf* dst, CompBuf* src)
{
// colormodulation and scale factors (cm & scalef) for 16 passes max: 64
int x, y, n, p, np;
fRGB c, tc, cm[64];
float sc, isc, u, v, sm, s, t, ofs, scalef[64];
CompBuf *tbuf1, *tbuf2, *gbuf;
const float cmo = 1.f - ndg->colmod;
const int qt = 1 << ndg->quality;
const float s1 = 4.f/(float)qt, s2 = 2.f*s1;
gbuf = BTP(src, ndg->threshold, qt);
tbuf1 = dupalloc_compbuf(gbuf);
IIR_gauss(tbuf1, s1, 0, 3);
IIR_gauss(tbuf1, s1, 1, 3);
IIR_gauss(tbuf1, s1, 2, 3);
tbuf2 = dupalloc_compbuf(tbuf1);
IIR_gauss(tbuf2, s2, 0, 3);
IIR_gauss(tbuf2, s2, 1, 3);
IIR_gauss(tbuf2, s2, 2, 3);
if (ndg->iter & 1) ofs = 0.5f; else ofs = 0.f;
for (x=0; x<(ndg->iter*4); x++) {
y = x & 3;
cm[x][0] = cm[x][1] = cm[x][2] = 1;
if (y==1) fRGB_rgbmult(cm[x], 1.f, cmo, cmo);
if (y==2) fRGB_rgbmult(cm[x], cmo, cmo, 1.f);
if (y==3) fRGB_rgbmult(cm[x], cmo, 1.f, cmo);
scalef[x] = 2.1f*(1.f-(x+ofs)/(float)(ndg->iter*4));
if (x & 1) scalef[x] = -0.99f/scalef[x];
}
sc = 2.13;
isc = -0.97;
for (y=0; y<gbuf->y; y++) {
v = (float)(y+0.5f) / (float)gbuf->y;
for (x=0; x<gbuf->x; x++) {
u = (float)(x+0.5f) / (float)gbuf->x;
s = (u-0.5f)*sc + 0.5f, t = (v-0.5f)*sc + 0.5f;
qd_getPixelLerp(tbuf1, s*gbuf->x, t*gbuf->y, c);
sm = smoothMask(s, t);
fRGB_mult(c, sm);
s = (u-0.5f)*isc + 0.5f, t = (v-0.5f)*isc + 0.5f;
qd_getPixelLerp(tbuf2, s*gbuf->x - 0.5f, t*gbuf->y - 0.5f, tc);
sm = smoothMask(s, t);
fRGB_madd(c, tc, sm);
qd_setPixel(gbuf, x, y, c);
}
}
memset(tbuf1->rect, 0, tbuf1->x*tbuf1->y*tbuf1->type*sizeof(float));
for (n=1; n<ndg->iter; n++) {
for (y=0; y<gbuf->y; y++) {
v = (float)(y+0.5f) / (float)gbuf->y;
for (x=0; x<gbuf->x; x++) {
u = (float)(x+0.5f) / (float)gbuf->x;
tc[0] = tc[1] = tc[2] = 0.f;
for (p=0;p<4;p++) {
np = (n<<2) + p;
s = (u-0.5f)*scalef[np] + 0.5f;
t = (v-0.5f)*scalef[np] + 0.5f;
qd_getPixelLerp(gbuf, s*gbuf->x - 0.5f, t*gbuf->y - 0.5f, c);
fRGB_colormult(c, cm[np]);
sm = smoothMask(s, t)*0.25f;
fRGB_madd(tc, c, sm);
}
p = (x + y*tbuf1->x)*tbuf1->type;
tbuf1->rect[p] += tc[0];
tbuf1->rect[p+1] += tc[1];
tbuf1->rect[p+2] += tc[2];
}
}
memcpy(gbuf->rect, tbuf1->rect, tbuf1->x*tbuf1->y*tbuf1->type*sizeof(float));
}
free_compbuf(tbuf1);
free_compbuf(tbuf2);
mixImages(dst, gbuf, 0.5f + 0.5f*ndg->mix);
free_compbuf(gbuf);
}
static void star4(NodeGlare* ndg, CompBuf* dst, CompBuf* src)
{
int x, y, i, xm, xp, ym, yp;
float c[4] = {0,0,0,0}, tc[4] = {0,0,0,0};
CompBuf *tbuf1, *tbuf2, *tsrc;
const float f1 = 1.f - ndg->fade, f2 = (1.f - f1)*0.5f;
//const float t3 = ndg->threshold*3.f;
const float sc = (float)(1 << ndg->quality);
const float isc = 1.f/sc;
tsrc = BTP(src, ndg->threshold, (int)sc);
tbuf1 = dupalloc_compbuf(tsrc);
tbuf2 = dupalloc_compbuf(tsrc);
for (i=0; i<ndg->iter; i++) {
// (x || x-1, y-1) to (x || x+1, y+1)
// F
for (y=0; y<tbuf1->y; y++) {
ym = y - i;
yp = y + i;
for (x=0; x<tbuf1->x; x++) {
xm = x - i;
xp = x + i;
qd_getPixel(tbuf1, x, y, c);
fRGB_mult(c, f1);
qd_getPixel(tbuf1, (ndg->angle ? xm : x), ym, tc);
fRGB_madd(c, tc, f2);
qd_getPixel(tbuf1, (ndg->angle ? xp : x), yp, tc);
fRGB_madd(c, tc, f2);
qd_setPixel(tbuf1, x, y, c);
}
}
// B
for (y=tbuf1->y-1; y>=0; y--) {
ym = y - i;
yp = y + i;
for (x=tbuf1->x-1; x>=0; x--) {
xm = x - i;
xp = x + i;
qd_getPixel(tbuf1, x, y, c);
fRGB_mult(c, f1);
qd_getPixel(tbuf1, (ndg->angle ? xm : x), ym, tc);
fRGB_madd(c, tc, f2);
qd_getPixel(tbuf1, (ndg->angle ? xp : x), yp, tc);
fRGB_madd(c, tc, f2);
qd_setPixel(tbuf1, x, y, c);
}
}
// (x-1, y || y+1) to (x+1, y || y-1)
// F
for (y=0; y<tbuf2->y; y++) {
ym = y - i;
yp = y + i;
for (x=0; x<tbuf2->x; x++) {
xm = x - i;
xp = x + i;
qd_getPixel(tbuf2, x, y, c);
fRGB_mult(c, f1);
qd_getPixel(tbuf2, xm, (ndg->angle ? yp : y), tc);
fRGB_madd(c, tc, f2);
qd_getPixel(tbuf2, xp, (ndg->angle ? ym : y), tc);
fRGB_madd(c, tc, f2);
qd_setPixel(tbuf2, x, y, c);
}
}
// B
for (y=tbuf2->y-1; y>=0; y--) {
ym = y - i;
yp = y + i;
for (x=tbuf2->x-1; x>=0; x--) {
xm = x - i;
xp = x + i;
qd_getPixel(tbuf2, x, y, c);
fRGB_mult(c, f1);
qd_getPixel(tbuf2, xm, (ndg->angle ? yp : y), tc);
fRGB_madd(c, tc, f2);
qd_getPixel(tbuf2, xp, (ndg->angle ? ym : y), tc);
fRGB_madd(c, tc, f2);
qd_setPixel(tbuf2, x, y, c);
}
}
}
for (y=0; y<tbuf1->y; ++y)
for (x=0; x<tbuf1->x; ++x) {
unsigned int p = (x + y*tbuf1->x)*tbuf1->type;
tbuf1->rect[p] += tbuf2->rect[p];
tbuf1->rect[p+1] += tbuf2->rect[p+1];
tbuf1->rect[p+2] += tbuf2->rect[p+2];
}
for (y=0; y<dst->y; ++y) {
const float m = 0.5f + 0.5f*ndg->mix;
for (x=0; x<dst->x; ++x) {
unsigned int p = (x + y*dst->x)*dst->type;
qd_getPixelLerp(tbuf1, x*isc, y*isc, tc);
dst->rect[p] = src->rect[p] + m*(tc[0] - src->rect[p]);
dst->rect[p+1] = src->rect[p+1] + m*(tc[1] - src->rect[p+1]);
dst->rect[p+2] = src->rect[p+2] + m*(tc[2] - src->rect[p+2]);
}
}
free_compbuf(tbuf1);
free_compbuf(tbuf2);
free_compbuf(tsrc);
}
static void lensDistort(CompBuf* dst, CompBuf* src, float kr, float kg, float kb, int jit, int proj, int fit)
{
int x, y, z;
const float cx = 0.5f*(float)dst->x, cy = 0.5f*(float)dst->y;
if (proj) {
// shift
CompBuf* tsrc = dupalloc_compbuf(src);
for (z=0; z<tsrc->type; ++z)
IIR_gauss(tsrc, (kr+0.5f)*(kr+0.5f), z, 1);
kr *= 20.f;
for (y=0; y<dst->y; y++) {
fRGB* colp = (fRGB*)&dst->rect[y*dst->x*dst->type];
const float v = (y + 0.5f)/(float)dst->y;
for (x=0; x<dst->x; x++) {
const float u = (x + 0.5f)/(float)dst->x;
qd_getPixelLerpChan(tsrc, (u*dst->x + kr) - 0.5f, v*dst->y - 0.5f, 0, colp[x]);
if (tsrc->type == CB_VAL)
colp[x][1] = tsrc->rect[x + y*tsrc->x];
else
colp[x][1] = tsrc->rect[(x + y*tsrc->x)*tsrc->type + 1];
qd_getPixelLerpChan(tsrc, (u*dst->x - kr) - 0.5f, v*dst->y - 0.5f, 2, colp[x]+2);
}
}
free_compbuf(tsrc);
}
else {
// Spherical
// Scale factor to make bottom/top & right/left sides fit in window after deform
// so in the case of pincushion (kn < 0), corners will be outside window.
// Now also optionally scales image such that black areas are not visible when distort factor is positive
// (makes distorted corners match window corners, but really only valid if mk<=0.5)
const float mk = MAX3(kr, kg, kb);
const float sc = (fit && (mk > 0.f)) ? (1.f/(1.f + 2.f*mk)) : (1.f/(1.f + mk));
const float drg = 4.f*(kg - kr), dgb = 4.f*(kb - kg);
kr *= 4.f, kg *= 4.f, kb *= 4.f;
for (y=0; y<dst->y; y++) {
fRGB* colp = (fRGB*)&dst->rect[y*dst->x*dst->type];
const float v = sc*((y + 0.5f) - cy)/cy;
for (x=0; x<dst->x; x++) {
int dr = 0, dg = 0, db = 0;
float d, t, ln[6] = {0, 0, 0, 0, 0, 0};
fRGB c1, tc = {0, 0, 0, 0};
const float u = sc*((x + 0.5f) - cx)/cx;
int sta = 0, mid = 0, end = 0;
if ((t = 1.f - kr*(u*u + v*v)) >= 0.f) {
d = 1.f/(1.f + sqrtf(t));
ln[0] = (u*d + 0.5f)*dst->x - 0.5f, ln[1] = (v*d + 0.5f)*dst->y - 0.5f;
sta = 1;
}
if ((t = 1.f - kg*(u*u + v*v)) >= 0.f) {
d = 1.f/(1.f + sqrtf(t));
ln[2] = (u*d + 0.5f)*dst->x - 0.5f, ln[3] = (v*d + 0.5f)*dst->y - 0.5f;
mid = 1;
}
if ((t = 1.f - kb*(u*u + v*v)) >= 0.f) {
d = 1.f/(1.f + sqrtf(t));
ln[4] = (u*d + 0.5f)*dst->x - 0.5f, ln[5] = (v*d + 0.5f)*dst->y - 0.5f;
end = 1;
}
if (sta && mid && end) {
// RG
const int dx = ln[2] - ln[0], dy = ln[3] - ln[1];
const float dsf = sqrtf(dx*dx + dy*dy) + 1.f;
const int ds = (int)(jit ? ((dsf < 4.f) ? 2.f : sqrtf(dsf)) : dsf);
const float sd = 1.f/(float)ds;
for (z=0; z<ds; ++z) {
const float tz = ((float)z + (jit ? BLI_frand() : 0.5f))*sd;
t = 1.f - (kr + tz*drg)*(u*u + v*v);
d = 1.f / (1.f + sqrtf(t));
qd_getPixelLerp(src, (u*d + 0.5f)*dst->x - 0.5f, (v*d + 0.5f)*dst->y - 0.5f, c1);
if (src->type == CB_VAL) c1[1] = c1[2] = c1[0];
tc[0] += (1.f-tz)*c1[0], tc[1] += tz*c1[1];
dr++, dg++;
}
// GB
{
const int dx = ln[4] - ln[2], dy = ln[5] - ln[3];
const float dsf = sqrtf(dx*dx + dy*dy) + 1.f;
const int ds = (int)(jit ? ((dsf < 4.f) ? 2.f : sqrtf(dsf)) : dsf);
const float sd = 1.f/(float)ds;
for (z=0; z<ds; ++z) {
const float tz = ((float)z + (jit ? BLI_frand() : 0.5f))*sd;
t = 1.f - (kg + tz*dgb)*(u*u + v*v);
d = 1.f / (1.f + sqrtf(t));
qd_getPixelLerp(src, (u*d + 0.5f)*dst->x - 0.5f, (v*d + 0.5f)*dst->y - 0.5f, c1);
if (src->type == CB_VAL) c1[1] = c1[2] = c1[0];
tc[1] += (1.f-tz)*c1[1], tc[2] += tz*c1[2];
dg++, db++;
}
}
}
if (dr) colp[x][0] = 2.f*tc[0] / (float)dr;
if (dg) colp[x][1] = 2.f*tc[1] / (float)dg;
if (db) colp[x][2] = 2.f*tc[2] / (float)db;
}
}
}
}