Update of /cvsroot/ufraw/ufraw
In directory sfp-cvsdas-2.v30.ch3.sourceforge.com:/tmp/cvs-serv15652
Modified Files:
ufraw.h ufraw_ufraw.c
Log Message:
Use bi-linear interpolation instead of Lanczos interpolation. Lanczos gives
nice sharp results, but there is no control over the sharpening.
Index: ufraw.h
===================================================================
RCS file: /cvsroot/ufraw/ufraw/ufraw.h,v
retrieving revision 1.144
retrieving revision 1.145
diff -u -d -r1.144 -r1.145
--- ufraw.h 27 Jan 2010 21:23:22 -0000 1.144
+++ ufraw.h 30 Jan 2010 00:48:52 -0000 1.145
@@ -328,7 +328,6 @@
lfModifier *TCAmodifier;
lfModifier *modifier;
#endif /* HAVE_LENSFUN */
- void *lanczos_func; /* the Lanczos kernel */
int hotpixels;
gboolean mark_hotpixels;
unsigned raw_multiplier;
Index: ufraw_ufraw.c
===================================================================
RCS file: /cvsroot/ufraw/ufraw/ufraw_ufraw.c,v
retrieving revision 1.232
retrieving revision 1.233
diff -u -d -r1.232 -r1.233
--- ufraw_ufraw.c 29 Jan 2010 01:43:20 -0000 1.232
+++ ufraw_ufraw.c 30 Jan 2010 00:48:52 -0000 1.233
@@ -264,7 +264,6 @@
uf->modFlags = 0;
uf->TCAmodifier = NULL;
uf->modifier = NULL;
- uf->lanczos_func = NULL;
#endif
ufraw_message(UFRAW_SET_LOG, "ufraw_open: w:%d h:%d curvesize:%d\n",
raw->width, raw->height, raw->toneCurveSize);
@@ -693,7 +692,6 @@
lf_camera_destroy(uf->conf->camera);
lf_modifier_destroy(uf->modifier);
#endif
- g_free(uf->lanczos_func);
if ( uf->conf->darkframe!=NULL ) {
ufraw_close(uf->conf->darkframe);
g_free(uf->conf->darkframe);
@@ -790,28 +788,6 @@
return UFRAW_SUCCESS;
}
-/* Lanczos kernel is precomputed in a table with this resolution
- * The value below seems to be enough for HQ upscaling up to eight times
- */
-#define LANCZOS_TABLE_RES 256
-/* A support of 3 gives an overall sharper looking image, but
- * it is a) slower b) gives more sharpening artefacts
- */
-#define LANCZOS_SUPPORT 2
-/* Define this to use a floating-point implementation of Lanczos interpolation.
- * The integer implementation is a little bit less accurate, but MUCH faster
- * (even on machines with FPU - ~2.5 times faster on Core2); besides, it will
- * run a hell lot faster on computers without a FPU (e.g. PDAs).
- */
-//#define LANCZOS_DATA_FLOAT
-#ifdef LANCZOS_DATA_FLOAT
-#define LANCZOS_DATA_TYPE float
-#define LANCZOS_DATA_ONE 1.0
-#else
-#define LANCZOS_DATA_TYPE int
-#define LANCZOS_DATA_ONE 4096
-#endif
-
#ifdef HAVE_LENSFUN
static void ufraw_convert_image_vignetting(ufraw_data *uf,
ufraw_image_data *img, UFRectangle *area)
@@ -829,11 +805,9 @@
static void ufraw_convert_image_transform(ufraw_data *uf, ufraw_image_data
*img,
ufraw_image_data *outimg, UFRectangle *area)
{
- double sine = sin(uf->conf->rotationAngle * 2 * M_PI / 360);
- double cosine = cos(uf->conf->rotationAngle * 2 * M_PI / 360);
+ float sine = sin(uf->conf->rotationAngle * 2 * M_PI / 360);
+ float cosine = cos(uf->conf->rotationAngle * 2 * M_PI / 360);
- /* Use precomputed Lanczos kernel */
- LANCZOS_DATA_TYPE *lanczos_func = uf->lanczos_func;
int x, y, c;
#ifdef HAVE_LENSFUN
@@ -848,93 +822,39 @@
float srcX = srcX0 + x*cosine;
float srcY = srcY0 - x*sine;
#ifdef HAVE_LENSFUN
- float buff[2];
if (applyLF) {
+ float buff[2];
lf_modifier_apply_geometry_distortion(uf->modifier,
srcX, srcY, 1, 1, buff);
srcX = buff[0];
srcY = buff[1];
}
#endif
-#ifdef LANCZOS_DATA_FLOAT
- float xs = ceilf(srcX) - LANCZOS_SUPPORT;
- float xe = floorf(srcX) + LANCZOS_SUPPORT;
- float ys = ceilf(srcY) - LANCZOS_SUPPORT;
- float ye = floorf(srcY) + LANCZOS_SUPPORT;
- if (xs < 0 || ys < 0 || xe >= img->width || ye >= img->height) {
+ gint32 xx = (gint32)srcX;
+ gint32 yy = (gint32)srcY;
+ // TODO: better handling of the borders.
+ if (xx < 0 || yy < 0 || xx+1 >= img->width || yy+1 >= img->height) {
for (c = 0; c < 3; c++)
cur[c] = 0;
continue;
}
- int dsrc = img->width - (xe - xs) - 1;
-
- for (c = 0; c < 3; c++) {
- ufraw_image_type *src = (ufraw_image_type *)(img->buffer +
- (long)ys * img->rowstride + (long)xs * img->depth);
- float norm = 0.0;
- float sum = 0.0;
-
- float _dx = srcX - xs;
- float yc, dy = srcY - ys;
- for (yc = ys; yc <= ye; yc += 1.0, dy -= 1.0) {
- float xc, dx = _dx;
- for (xc = xs; xc <= xe; xc += 1.0, dx -= 1.0, src++) {
- float d = dx * dx + dy * dy;
- if (d >= LANCZOS_SUPPORT * LANCZOS_SUPPORT)
- continue;
-
- d = lanczos_func[(int)(d * LANCZOS_TABLE_RES)];
- norm += d;
- sum += d * src[0][c];
- }
- src += dsrc;
- }
- if (norm != 0.0)
- cur[c] = LIM((int)(sum / norm), 0, 0xffff);
- else
- cur[c] = 0;
- }
+ ufraw_image_type *src = (ufraw_image_type *)(img->buffer +
+ yy * img->rowstride + xx * img->depth);
+#if 1
+ /* Do it in integer arithmetic, it's a bit faster */
+ guint64 dx = (gint32)(srcX * 4096.0) - (xx << 12);
+ guint64 dy = (gint32)(srcY * 4096.0) - (yy << 12);
+ for (c = 0; c < 3; c++)
+ cur[c] = ( (4096-dy)*((4096-dx)*src[0][c] + dx*src[1][c])
+ + dy*((4096-dx)*src[img->width][c]
+ + (dx)*src[img->width+1][c]) ) >> 24;
#else
- /* Do it in integer arithmetic, it's faster */
- int xx = (int)srcX;
- int yy = (int)srcY;
- int xs = xx + 1 - LANCZOS_SUPPORT;
- int xe = xx + LANCZOS_SUPPORT;
- int ys = yy + 1 - LANCZOS_SUPPORT;
- int ye = yy + LANCZOS_SUPPORT;
- if (xs < 0 || ys < 0 || xe >= img->width || ye >= img->height) {
- for (c = 0; c < 3; c++)
- cur[c] = 0;
- continue;
- }
- int dsrc = img->width - (xe - xs) - 1;
-
- for (c = 0; c < 3; c++) {
- ufraw_image_type *src = (ufraw_image_type *)(img->buffer +
- ys * img->rowstride + xs * img->depth);
- int norm = 0;
- int sum = 0;
-
- int _dx = (int)(srcX * 4096.0) - (xs << 12);
- int yc, dy = (int)(srcY * 4096.0) - (ys << 12);
- for (yc = ys; yc <= ye; yc++, dy -= 4096) {
- int xc, dx = _dx;
- for (xc = xs; xc <= xe; xc++, src++, dx -= 4096) {
- int d = (dx * dx + dy * dy) >> 12;
- if (d >= 4096 * LANCZOS_SUPPORT * LANCZOS_SUPPORT)
- continue;
-
- d = lanczos_func[(d * LANCZOS_TABLE_RES) >> 12];
- norm += d;
- sum += d * src[0][c];
- }
- src += dsrc;
- }
- if (norm != 0)
- cur[c] = LIM(sum / norm, 0, 0xffff);
- else
- cur[c] = 0;
- }
+ float dx = srcX - xx;
+ float dy = srcY - yy;
+ for (c = 0; c < 3; c++)
+ cur[c] = (1-dy)*((1-dx)*src[0][c] + dx*src[1][c])
+ + dy*((1-dx)*src[img->width][c]
+ + (dx)*src[img->width+1][c]);
#endif
}
}
@@ -1545,25 +1465,6 @@
static void ufraw_prepare_transform(ufraw_data *uf)
{
- if (uf->lanczos_func == NULL) {
- /* Precompute the Lanczos kernel */
- LANCZOS_DATA_TYPE *lanczos_func = g_new(LANCZOS_DATA_TYPE,
- LANCZOS_SUPPORT * LANCZOS_SUPPORT * LANCZOS_TABLE_RES);
- uf->lanczos_func = lanczos_func;
- int i;
- for (i = 0; i < LANCZOS_SUPPORT*LANCZOS_SUPPORT * LANCZOS_TABLE_RES;
- i++) {
- if (i == 0) {
- lanczos_func[i] = LANCZOS_DATA_ONE;
- } else {
- float d = sqrt((float)i / LANCZOS_TABLE_RES);
- lanczos_func[i] = (LANCZOS_DATA_TYPE)(
- LANCZOS_DATA_ONE * LANCZOS_SUPPORT *
- sin(M_PI * d) * sin(M_PI / LANCZOS_SUPPORT * d) /
- (M_PI * M_PI * d * d) );
- }
- }
- }
#ifdef HAVE_LENSFUN
ufraw_image_data *img = &uf->Images[ufraw_first_phase];
conf_data *conf = uf->conf;
@@ -1589,6 +1490,8 @@
lf_modifier_destroy(uf->modifier);
uf->modifier = NULL;
}
+#else /* HAVE_LENSFUN */
+ (void)uf;
#endif /* HAVE_LENSFUN */
}
------------------------------------------------------------------------------
The Planet: dedicated and managed hosting, cloud storage, colocation
Stay online with enterprise data centers and the best network in the business
Choose flexible plans and management services without long-term contracts
Personal 24x7 support from experience hosting pros just a phone call away.
http://p.sf.net/sfu/theplanet-com
_______________________________________________
ufraw-cvs mailing list
ufraw-cvs@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ufraw-cvs
