blender
/
blender
129
397
Fork 571
Code Issues 6.2k Pull Requests 750 Projects 19 Wiki Activity

Cycles Denoising: Skip confidence interval test for outlier central pixels 

If the central pixel is an outlier, the denoiser is supposed to predict its
value from the surrounding pixels. However, in some cases the confidence
interval test would reject every single surrounding pixel, which leaves the
model fitting with no data to work with.
This commit is contained in:
Lukas Stockner 2017-05-21 05:26:13 +02:00
parent 38a2bf665b
commit 96769f3b19
1 changed files with 8 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
intern/cycles/kernel/filter/filter_reconstruction.h
View File

@ -55,9 +55,14 @@ ccl_device_inline void kernel_filter_construct_gramian(int x, int y,
float q_std_dev = sqrtf(filter_get_pixel_variance(variance_pass + q_offset, pass_stride));
/* If the pixel was flagged as an outlier during prefiltering, skip it.
* Otherwise, perform the regular confidence interval test. */
if(ccl_get_feature(buffer + q_offset, 0) < 0.0f ||
average(fabs(p_color - q_color)) > 2.0f*(p_std_dev + q_std_dev + 1e-3f)) {
* Otherwise, perform the regular confidence interval test unless
* the center pixel is an outlier (in that case, using the confidence
* interval test could result in no pixels being used at all). */
bool p_outlier = (ccl_get_feature(buffer + p_offset, 0) < 0.0f);
bool q_outlier = (ccl_get_feature(buffer + q_offset, 0) < 0.0f);
bool outside_of_interval = (average(fabs(p_color - q_color)) > 2.0f*(p_std_dev + q_std_dev + 1e-3f));
if(q_outlier || (!p_outlier && outside_of_interval)) {
return;
}