| 17 |
|
* each of NI2DIR surrounding directions. To speed this |
| 18 |
|
* calculation, we sort the data into half-planes and apply |
| 19 |
|
* simple tests to see which neighbor is closest in each |
| 20 |
< |
* direction. Once we have our approximate neighborhood |
| 20 |
> |
* angular slice. Once we have our approximate neighborhood |
| 21 |
|
* for a sample, we can use it in a modified Gaussian weighting |
| 22 |
|
* with allowing local anisotropy. Harmonic weighting is added |
| 23 |
|
* to reduce the influence of distant neighbors. This yields a |
| 71 |
|
interp2_free(ip); |
| 72 |
|
return(NULL); |
| 73 |
|
} |
| 74 |
< |
if (nsamps == ip->ns); |
| 74 |
> |
if (nsamps == ip->ns) |
| 75 |
|
return(ip); |
| 76 |
|
if (ip->da != NULL) { /* will need to recompute distribution */ |
| 77 |
|
free(ip->da); |
| 90 |
|
{ |
| 91 |
|
if (mind <= 0) |
| 92 |
|
return; |
| 93 |
< |
if ((.998*ip->dmin <= mind) && (mind <= 1.002*ip->dmin)) |
| 93 |
> |
if ((.998*ip->dmin <= mind) & (mind <= 1.002*ip->dmin)) |
| 94 |
|
return; |
| 95 |
|
if (ip->da != NULL) { /* will need to recompute distribution */ |
| 96 |
|
free(ip->da); |
| 303 |
|
/* identify top n weights */ |
| 304 |
|
for (i = ip->ns; i--; ) { |
| 305 |
|
const double wti = get_wt(ip, i, x, y); |
| 306 |
+ |
if (wti <= 1e-9) |
| 307 |
+ |
continue; |
| 308 |
|
for (j = nn; j > 0; j--) { |
| 309 |
|
if (wt[j-1] >= wti) |
| 310 |
|
break; |
