diff options
author | Matt Ebb <matt@mke3.net> | 2010-05-26 12:06:51 +0400 |
---|---|---|
committer | Matt Ebb <matt@mke3.net> | 2010-05-26 12:06:51 +0400 |
commit | 440917cb0b2fe7f736d5911eb40318a0bfc8f0d4 (patch) | |
tree | 6bc197fa279d85ce6f936a02db54c9a39cae4489 /source/blender/nodes | |
parent | 244cda6f1ae3a0311c118953197704f1391eb702 (diff) |
Fix [#22418] Displace Node also makes entire image fuzzy
Changed displace node sampling to use EWA filtering, and removed old hacks
for calculating derivatives - I think it should be generated correctly now.
Diffstat (limited to 'source/blender/nodes')
-rw-r--r-- | source/blender/nodes/intern/CMP_nodes/CMP_displace.c | 135 |
1 files changed, 62 insertions, 73 deletions
diff --git a/source/blender/nodes/intern/CMP_nodes/CMP_displace.c b/source/blender/nodes/intern/CMP_nodes/CMP_displace.c index 877f71bfbfb..6fe6dcd8440 100644 --- a/source/blender/nodes/intern/CMP_nodes/CMP_displace.c +++ b/source/blender/nodes/intern/CMP_nodes/CMP_displace.c @@ -44,94 +44,83 @@ static bNodeSocketType cmp_node_displace_out[]= { { -1, 0, "" } }; -static float *vecbuf_get_pixel(CompBuf *vecbuf, float *veccol, int x, int y) -{ - /* the x-xrad stuff is a bit weird, but i seem to need it otherwise - * my returned pixels are offset weirdly */ - return compbuf_get_pixel(vecbuf, veccol, x-vecbuf->xrad, y-vecbuf->yrad, vecbuf->xrad, vecbuf->yrad); -} +/* minimum distance (in pixels) a pixel has to be displaced + * in order to take effect */ +#define DISPLACE_EPSILON 0.01 static void do_displace(CompBuf *stackbuf, CompBuf *cbuf, CompBuf *vecbuf, float *veccol, float *xscale, float *yscale) { ImBuf *ibuf; - float dx=0.0, dy=0.0; - float dspx, dspy; - float uv[2], col[4], colnext[4], colprev[4]; - float *vp, *vpnext, *vpprev; - float *out= stackbuf->rect, *vec=vecbuf->rect, *in= cbuf->rect; - int x, y, vx, vy, sx, sy; + int x, y; + float p_dx, p_dy; /* main displacement in pixel space */ + float d_dx, d_dy; + float dxt, dyt; + float u, v; + float vec[3], vecdx[3], vecdy[3]; + float col[3]; - /* ibuf needed for sampling */ ibuf= IMB_allocImBuf(cbuf->x, cbuf->y, 32, 0, 0); ibuf->rect_float= cbuf->rect; - vec = vecbuf->rect; + for(y=0; y < stackbuf->y; y++) { + for(x=0; x < stackbuf->x; x++) { + /* calc pixel coordinates */ + qd_getPixel(vecbuf, x-vecbuf->xof, y-vecbuf->yof, vec); + p_dx = vec[0] * xscale[0]; + p_dy = vec[1] * yscale[0]; + + /* if no displacement, then just copy this pixel */ + if (p_dx < DISPLACE_EPSILON && p_dy < DISPLACE_EPSILON) { + qd_getPixel(cbuf, x-cbuf->xof, y-cbuf->yof, col); + qd_setPixel(stackbuf, x, y, col); + continue; + } + + /* displaced pixel in uv coords, for image sampling */ + u = (x - cbuf->xof - p_dx + 0.5f) / (float)stackbuf->x; + v = (y - cbuf->yof - p_dy + 0.5f) / (float)stackbuf->y; + + + /* calc derivatives */ + qd_getPixel(vecbuf, x-vecbuf->xof+1, y-vecbuf->yof, vecdx); + qd_getPixel(vecbuf, x-vecbuf->xof, y-vecbuf->yof+1, vecdy); + d_dx = vecdx[0] * xscale[0]; + d_dy = vecdy[0] * yscale[0]; + + /* clamp derivatives to minimum displacement distance in UV space */ + dxt = MAX2(p_dx - d_dx, DISPLACE_EPSILON)/(float)stackbuf->x; + dyt = MAX2(p_dy - d_dy, DISPLACE_EPSILON)/(float)stackbuf->y; + + ibuf_sample(ibuf, u, v, dxt, dyt, col); + qd_setPixel(stackbuf, x, y, col); + } + } + IMB_freeImBuf(ibuf); - sx= stackbuf->x; - sy= stackbuf->y; - - QUATCOPY(col, veccol); - QUATCOPY(colnext, veccol); - QUATCOPY(colprev, veccol); - for(y=0; y<sy; y++) { - for(x= 0; x< sx; x++, out+=4, in+=4, vec+=3) { +/* simple method for reference, linear interpolation */ +/* + int x, y; + float dx, dy; + float u, v; + float vec[3]; + float col[3]; + + for(y=0; y < stackbuf->y; y++) { + for(x=0; x < stackbuf->x; x++) { + qd_getPixel(vecbuf, x, y, vec); - vp = vecbuf_get_pixel(vecbuf, col, x, y); - - /* this happens in compbuf_get_pixel, need to make sure the following - * check takes them into account */ - vx= x-vecbuf->xof; - vy= y-vecbuf->yof; + dx = vec[0] * (xscale[0]); + dy = vec[1] * (yscale[0]); - /* find the new displaced co-ords, also correcting for translate offset */ - dspx = vx - (*xscale * vp[0]); - dspy = vy - (*yscale * vp[1]); - - /* convert image space to 0.0-1.0 UV space for sampling, correcting for translate offset */ - uv[0] = dspx / (float)sx; - uv[1] = dspy / (float)sy; - - if(vx>0 && vx< vecbuf->x-1 && vy>0 && vy< vecbuf->y-1) { - /* adaptive sampling, X and Y channel. - * we call vecbuf_get_pixel for every pixel since the input - * might be a procedural, and then we can't use offsets */ - vpprev = vecbuf_get_pixel(vecbuf, colprev, x-1, y); - vpnext = vecbuf_get_pixel(vecbuf, colnext, x+1, y); - dx= 0.5f*(fabs(vp[0]-vpprev[0]) + fabs(vp[0]-vpnext[0])); - - vpprev = vecbuf_get_pixel(vecbuf, colprev, x, y-1); - vpnext = vecbuf_get_pixel(vecbuf, colnext, x, y+1); - dy= 0.5f*(fabs(vp[1]-vpnext[1]) + fabs(vp[1]-vpprev[1])); - - vpprev = vecbuf_get_pixel(vecbuf, colprev, x-1, y-1); - vpnext = vecbuf_get_pixel(vecbuf, colnext, x-1, y+1); - dx+= 0.25f*(fabs(vp[0]-vpprev[0]) + fabs(vp[0]-vpnext[0])); - dy+= 0.25f*(fabs(vp[1]-vpprev[1]) + fabs(vp[1]-vpnext[1])); - - vpprev = vecbuf_get_pixel(vecbuf, colprev, x+1, y-1); - vpnext = vecbuf_get_pixel(vecbuf, colnext, x+1, y+1); - dx+= 0.25f*(fabs(vp[0]-vpprev[0]) + fabs(vp[0]-vpnext[0])); - dy+= 0.25f*(fabs(vp[1]-vpprev[1]) + fabs(vp[1]-vpnext[1])); - - /* scaled down to prevent blurriness */ - /* 8: magic number, provides a good level of sharpness without getting too aliased */ - dx /= 8; - dy /= 8; - } - - /* should use mipmap */ - if(dx > 0.006f) dx= 0.006f; - if(dy > 0.006f) dy= 0.006f; - if ((vp[0]> 0.0) && (dx < 0.004)) dx = 0.004; - if ((vp[1]> 0.0) && (dy < 0.004)) dy = 0.004; + u = (x - dx + 0.5f) / (float)stackbuf->x; + v = (y - dy + 0.5f) / (float)stackbuf->y; - - ibuf_sample(ibuf, uv[0], uv[1], dx, dy, out); + qd_getPixelLerp(cbuf, u*cbuf->x - 0.5f, v*cbuf->y - 0.5f, col); + qd_setPixel(stackbuf, x, y, col); } } - - IMB_freeImBuf(ibuf); +*/ } |