sergiotarxz/libvips
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

improve the GIF save change detector

Browse Source 
Instead if just summing PELs, scale RGBA differently. This lets us
detect changes like [255, 0, 0] to [9, 255, 0]
This commit is contained in:
John Cupitt 2022-03-02 18:03:36 +00:00
parent de2adb99a2
commit 669ea43138
1 changed files with 16 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
libvips/foreign/cgifsave.c
View File

@ -91,12 +91,9 @@ typedef struct _VipsForeignSaveCgif {
const VipsQuantisePalette *lp;
/* The current colourmap, updated on a significant frame change.
*
* frame_sum is 32-bit, so we can handle a max of about 2000 x 2000
* RGB pixel per frame.
*/
VipsPel *palette_rgb;
guint frame_sum;
guint64 frame_sum;
/* The index frame we get libimagequant to generate.
*/
@ -193,14 +190,13 @@ vips_foreign_save_cgif_write_frame( VipsForeignSaveCgif *cgif )
/* We know this fits in an int since we limit frame size.
*/
int n_pels = frame_rect->height * frame_rect->width;
guint max_sum = 256 * n_pels * 4;
VipsPel *frame_bytes =
VIPS_REGION_ADDR( cgif->frame, 0, frame_rect->top );
VipsPel * restrict p;
VipsPel *rgb;
guint sum;
double percent_change;
guint64 sum;
double change;
int i;
CGIF_FrameConfig frame_config;
@ -230,14 +226,21 @@ vips_foreign_save_cgif_write_frame( VipsForeignSaveCgif *cgif )
*/
sum = 0;
p = frame_bytes;
for( i = 0; i < n_pels * 4; i++ )
sum += p[i];
percent_change = 100 *
fabs( ((double) sum / max_sum) -
((double) cgif->frame_sum / max_sum) );
for( i = 0; i < n_pels; i++ ) {
/* Scale RGBA differently so that changes like [0, 255, 0]
* to [255, 0, 0] are detected.
*/
sum += p[0] * 1000;
sum += p[1] * 100;
sum += p[2] * 10;
sum += p[3];
p += 4;
}
change = VIPS_ABS( (sum - cgif->frame_sum) / n_pels );
if( cgif->frame_sum == 0 ||
percent_change > 0 ) {
change > 0 ) {
cgif->frame_sum = sum;
/* If this is not our first cmap, make a note that we need to