mirror-github-bspeice/cuburn
1
0
Fork 0
mirror of https://github.com/stevenrobertson/cuburn.git synced 2025-05-14 12:21:29 -04:00
Code Issues Actions Packages Projects Releases Wiki Activity

Expose options on filter

Browse Source
This commit is contained in:
Steven Robertson 2012-01-26 05:34:43 -05:00
parent a4b482f61a
commit 899b8082da
1 changed files with 9 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
cuburn/code/filtering.py
View File

@ -1,4 +1,3 @@
import numpy as np import numpy as np
from numpy import float32 as f32, int32 as i32 from numpy import float32 as f32, int32 as i32
@ -172,9 +171,12 @@ __global__ void blur_v(float *dst) {
dst[y * (blockDim.x * gridDim.x) + x] = den; dst[y * (blockDim.x * gridDim.x) + x] = den;
} }
/* sstd: spatial standard deviation (Gaussian filter) /* sstd: spatial standard deviation (Gaussian filter)
* cstd: color standard deviation (Gaussian on the range [0, 1], where 1 * cstd: color standard deviation (Gaussian on the range [0, 1], where 1
* represents an "opposite" color). * represents an "opposite" color).
* angstd: inverse standard deviation of negative of cosine of angle
* between current filter direction and density gradient direction
* (yes, this is absurd; no, I'm not joking)
* *
* Density is controlled by a power-of-two Gompertz distribution: * Density is controlled by a power-of-two Gompertz distribution:
* v = 1 - 2^(-sum^dpow * 2^((dhalfpt - x) * dspeed)) * v = 1 - 2^(-sum^dpow * 2^((dhalfpt - x) * dspeed))
@ -192,13 +194,11 @@ __global__ void blur_v(float *dst) {
* cell. * cell.
*/ */
__global__ __global__
void bilateral(float4 *dst, float sstd, float cstd, void bilateral(float4 *dst, int r, float sstd, float cstd, float angscale,
float dhalfpt, float dspeed, float dpow) { float dhalfpt, float dspeed, float dpow) {
int x = blockIdx.x * blockDim.x + threadIdx.x; int x = blockIdx.x * blockDim.x + threadIdx.x;
int y = blockIdx.y * blockDim.y + threadIdx.y; int y = blockIdx.y * blockDim.y + threadIdx.y;
const int r = 7;
// Precalculate the spatial coeffecients. // Precalculate the spatial coeffecients.
__shared__ float spa_coefs[32]; __shared__ float spa_coefs[32];
if (threadIdx.y == 0) { if (threadIdx.y == 0) {
@ -260,8 +260,7 @@ void bilateral(float4 *dst, float sstd, float cstd,
// Oh, this is ridiculous. But it works! // Oh, this is ridiculous. But it works!
float factor = spa_coefs[abs(i)] * spa_coefs[abs(j)] float factor = spa_coefs[abs(i)] * spa_coefs[abs(j)]
* expf(cscale * cdiff) * dfact * expf(cscale * cdiff) * dfact
* exp2f(2.0f * (-angfact - 1.0f)); * exp2f(angscale * (-angfact - 1.0f));
weightsum += factor; weightsum += factor;
out.x += factor * pix.x; out.x += factor * pix.x;
out.y += factor * pix.y; out.y += factor * pix.y;
@ -328,7 +327,7 @@ class Filtering(HunkOCode):
bilateral, blur_h, blur_h_1cp, blur_v, fma_buf = map( bilateral, blur_h, blur_h_1cp, blur_v, fma_buf = map(
self.mod.get_function, self.mod.get_function,
'bilateral blur_h blur_h_1cp blur_v fma_buf'.split()) 'bilateral blur_h blur_h_1cp blur_v fma_buf'.split())
ROUNDS = 2 # TODO: user customizable? ROUNDS = 1 # TODO: user customizable?
def do_bilateral(bsrc, bdst): def do_bilateral(bsrc, bdst):
tref.set_address_2d(bsrc, dsc, sb) tref.set_address_2d(bsrc, dsc, sb)