better hough_line

2x faster, better parameter space, but breaks compat :( sorry
This commit is contained in:
John Cupitt 2018-02-01 20:41:14 +00:00
parent f56fbf564f
commit 78db92e60e
2 changed files with 16 additions and 8 deletions

View File

@ -2,6 +2,8 @@
- add magicksave, save image with libMagick [dlemstra]
- remove jpeg thumbnail from EXIF if "jpeg-thumbnail-data" has been removed by
user
- hough_line scales width to 0 - 180, not 0 - 360
- hough_line is 2x faster
5/1/18 started 8.6.2
- vips_sink_screen() keeps a ref to the input image ... stops a rare race

View File

@ -2,6 +2,8 @@
*
* 7/3/14
* - from hist_find.c
* 1/2/18
* - change width to 0 - 180
*/
/*
@ -73,8 +75,10 @@ vips_hough_line_build( VipsObject *object )
if( !(hough_line->sin = VIPS_ARRAY( object, width, double )) )
return( -1 );
/* Map width to 180 degrees.
*/
for( i = 0; i < width; i++ )
hough_line->sin[i] = sin( 2 * VIPS_PI * i / width );
hough_line->sin[i] = sin( VIPS_PI * i / width );
if( VIPS_OBJECT_CLASS( vips_hough_line_parent_class )->build( object ) )
return( -1 );
@ -107,17 +111,19 @@ vips_hough_line_vote( VipsHough *hough, VipsImage *accumulator, int x, int y )
double yd = (double) y / statistic->ready->Ysize;
int width = hough_line->width;
int height = hough_line->height;
unsigned int *data =
(unsigned int *) VIPS_IMAGE_ADDR( accumulator, 0, 0 );
int i;
for( i = 0; i < width; i++ ) {
int i90 = (i + width / 4) % width;
for( i = 0; i < width; i++ ) {
int i90 = (i + width / 2) % width;
double r = xd * hough_line->sin[i90] + yd * hough_line->sin[i];
int ri = height * r;
if( ri >= 0 &&
ri < height )
*VIPS_IMAGE_ADDR( accumulator, i, ri ) += 1;
if( ri >= 0 &&
ri < height )
data[i + ri * width] += 1;
}
}
@ -174,8 +180,8 @@ vips_hough_line_init( VipsHoughLine *hough_line )
*
* Find the line Hough transform for @in. @in must have one band. @out has one
* band, with pixels being the number of votes for that line. The X dimension
* of @out is the line angle, the Y dimension is the distance of the line from
* the origin.
* of @out is the line angle in 0 - 180 degrees, the Y dimension is the
* distance of the closest part of that line to the origin in the top-left.
*
* Use @width @height to set the size of the parameter space image (@out),
* that is, how accurate the line determination should be.