| 243 |
|
ep[0] += d2; |
| 244 |
|
ep[-hp->ns] += d2; |
| 245 |
|
} |
| 246 |
< |
if (j) { /* from behind */ |
| 247 |
< |
d2 = b - bright(ap[-1].v); |
| 248 |
< |
d2 *= d2; |
| 249 |
< |
ep[0] += d2; |
| 250 |
< |
ep[-1] += d2; |
| 251 |
< |
} |
| 246 |
> |
if (!j) continue; |
| 247 |
> |
/* from behind */ |
| 248 |
> |
d2 = b - bright(ap[-1].v); |
| 249 |
> |
d2 *= d2; |
| 250 |
> |
ep[0] += d2; |
| 251 |
> |
ep[-1] += d2; |
| 252 |
> |
if (!i) continue; |
| 253 |
> |
/* diagonal */ |
| 254 |
> |
d2 = b - bright(ap[-hp->ns-1].v); |
| 255 |
> |
d2 *= d2; |
| 256 |
> |
ep[0] += d2; |
| 257 |
> |
ep[-hp->ns-1] += d2; |
| 258 |
|
} |
| 259 |
|
/* correct for number of neighbors */ |
| 260 |
< |
earr[0] *= 2.f; |
| 261 |
< |
earr[hp->ns-1] *= 2.f; |
| 262 |
< |
earr[(hp->ns-1)*hp->ns] *= 2.f; |
| 263 |
< |
earr[(hp->ns-1)*hp->ns + hp->ns-1] *= 2.f; |
| 260 |
> |
earr[0] *= 8./3.; |
| 261 |
> |
earr[hp->ns-1] *= 8./3.; |
| 262 |
> |
earr[(hp->ns-1)*hp->ns] *= 8./3.; |
| 263 |
> |
earr[(hp->ns-1)*hp->ns + hp->ns-1] *= 8./3.; |
| 264 |
|
for (i = 1; i < hp->ns-1; i++) { |
| 265 |
< |
earr[i*hp->ns] *= 4./3.; |
| 266 |
< |
earr[i*hp->ns + hp->ns-1] *= 4./3.; |
| 265 |
> |
earr[i*hp->ns] *= 8./5.; |
| 266 |
> |
earr[i*hp->ns + hp->ns-1] *= 8./5.; |
| 267 |
|
} |
| 268 |
|
for (j = 1; j < hp->ns-1; j++) { |
| 269 |
< |
earr[j] *= 4./3.; |
| 270 |
< |
earr[(hp->ns-1)*hp->ns + j] *= 4./3.; |
| 269 |
> |
earr[j] *= 8./5.; |
| 270 |
> |
earr[(hp->ns-1)*hp->ns + j] *= 8./5.; |
| 271 |
|
} |
| 272 |
|
return(earr); |
| 273 |
|
} |
| 278 |
|
ambsupersamp(double acol[3], AMBHEMI *hp, int cnt) |
| 279 |
|
{ |
| 280 |
|
float *earr = getambdiffs(hp); |
| 281 |
< |
double e2sum = 0.0; |
| 281 |
> |
double e2rem = 0; |
| 282 |
|
AMBSAMP *ap; |
| 283 |
|
RAY ar; |
| 284 |
|
double asum[3]; |
| 285 |
|
float *ep; |
| 286 |
< |
int i, j, n; |
| 286 |
> |
int i, j, n, nss; |
| 287 |
|
|
| 288 |
|
if (earr == NULL) /* just skip calc. if no memory */ |
| 289 |
|
return; |
| 290 |
< |
/* add up estimated variances */ |
| 291 |
< |
for (ep = earr + hp->ns*hp->ns; ep-- > earr; ) |
| 292 |
< |
e2sum += *ep; |
| 290 |
> |
/* accumulate estimated variances */ |
| 291 |
> |
for (ep = earr + hp->ns*hp->ns; ep > earr; ) |
| 292 |
> |
e2rem += *--ep; |
| 293 |
|
ep = earr; /* perform super-sampling */ |
| 294 |
|
for (ap = hp->sa, i = 0; i < hp->ns; i++) |
| 295 |
|
for (j = 0; j < hp->ns; j++, ap++) { |
| 296 |
< |
int nss = *ep/e2sum*cnt + frandom(); |
| 296 |
> |
if (e2rem <= FTINY) |
| 297 |
> |
goto done; /* nothing left to do */ |
| 298 |
> |
nss = *ep/e2rem*cnt + frandom(); |
| 299 |
|
asum[0] = asum[1] = asum[2] = 0.0; |
| 300 |
|
for (n = 1; n <= nss; n++) { |
| 301 |
|
if (!getambsamp(&ar, hp, i, j, n)) { |
| 305 |
|
addcolor(asum, ar.rcol); |
| 306 |
|
} |
| 307 |
|
if (nss) { /* update returned ambient value */ |
| 308 |
< |
const double ssf = 1./(nss + 1); |
| 308 |
> |
const double ssf = 1./(nss + 1.); |
| 309 |
|
for (n = 3; n--; ) |
| 310 |
|
acol[n] += ssf*asum[n] + |
| 311 |
|
(ssf - 1.)*colval(ap->v,n); |
| 312 |
|
} |
| 313 |
< |
e2sum -= *ep++; /* update remainders */ |
| 313 |
> |
e2rem -= *ep++; /* update remainders */ |
| 314 |
|
cnt -= nss; |
| 315 |
|
} |
| 316 |
+ |
done: |
| 317 |
|
free(earr); |
| 318 |
|
} |
| 319 |
|
|
