0x2620
/
pandora
2
1
Fork 1
Code Issues 1.3k Pull Requests Releases Wiki Activity

make cut detection faster

This commit is contained in:
rolux 2012-03-06 21:18:29 +01:00
parent c7856a70a6
commit 6f82e1298f
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
pandora/archive/extract.py
View File

@ -20,6 +20,7 @@ import ox.image
img_extension='jpg'
FFMPEG2THEORA = 'ffmpeg2theora'
MAX_DISTANCE = math.sqrt(3 * pow(255, 2))
class AspectRatio(fractions.Fraction):
@ -338,7 +339,7 @@ def average_volume(prefix, start=0, end=0):
def get_distance(rgb0, rgb1):
# rgb distance, normalized so that black/white equals 1
dst = math.sqrt(pow(rgb0[0] - rgb1[0], 2) + pow(rgb0[1] - rgb1[1], 2) + pow(rgb0[2] - rgb1[2], 2))
return dst / math.sqrt(3 * pow(255, 2))
return dst / MAX_DISTANCE
def cuts(prefix):
@ -364,8 +365,7 @@ def cuts(prefix):
rgb1 = pixels[image1][x, y]
distance += get_distance(rgb0, rgb1)
distances.append(distance / height)
for frame in range(1, frames):
if distances[frame] >= 0.025 and abs(distances[frame] - distances[frame - 1]) >= 0.05:
if distance >= 0.025 and abs(distance - distances[frame - 1]) >= 0.05:
cuts.append(frame / fps)
return cuts