[FFmpeg-devel] Possible concealment bug

Hardik Sharma hardik.sharma22 at yahoo.com
Sat Oct 8 06:04:00 CEST 2011


It is reproducible as I tried to drop different combination of slices and if I include slices from those 200 slices, it drop frames. My encoded stream is -
http://www.datafilehost.com/download-1c199ae3.html
The slice numbers which if you include in any random slice drop of any percentage are-
 
102   
63   
47   
53   
206   
72   
55   
208   
474   
81   
90   
65   
324   
83   
74   
191   
100   
622   
111   
92   
114   
125   
128   
199   
214   
193   
136   
332   
216   
360   
198   
227   
221   
220   
237   
235   
226   
247   
245   
341   
482   
349   
371   
255   
257   
326   
266   
632   
334   
351   
343   
380   
363   
373   
392   
491   
1065   
382   
501   
404   
394   
406   
476   
484   
415   
532   
511   
521   
493   
503   
513   
523   
643   
543   
535   
545   
555   
553   
719   
625   
1077   
1113   
656   
564   
668   
646   
693   
634   
659   
678   
970   
812   
1090   
670   
682   
706   
801   
946   
697   
731   
722   
933   
916   
944   
806   
733   
814   
924   
741   
832   
805   
956   
839   
834   
847   
821   
827   
820   
826   
999   
918   
841   
849   
935   
968   
926   
958   
855   
854   
860   
979   
1102   
992   
981   
990   
1001   
1010   
1068   
1080   
1123   
1093   
1146   
1133   
1105   
1116   
1159   
1125   
1136   
1181   
1165   
1261   
1255   
1186   
1322   
1199   
1270   
1263   
1278   
1272   
1286   
1276   
1298   
1284   
1295   
1326   
1307   
1342   
1361   
1339   
1414   
1364   
1350   
1419   
1426   
1434   
1443   
1451   
1459   
1468   
1475   
1483   
1492   
1606   
1501   
1553   
1620   
1549   
1511   
1520   
1600   
1529   
1539   
1613   
1562   
1635   
1627   
1648   
1654   
1642   
1661   
1670   
1681   
1854   
1692   
1860   
1738   
1701   
1868   
1879   
1711   
1721   
1740   
1890   
1605   
1763   
1862   
1474   
1870   
2084   
1619   
1881   
1466   
1482   
1612   
1892   
1913   
1458   
1914   
1929   
1633   
1928   
1940   
1941   
1951   
1952   
2095   
1964   
1491   
2066   
1101   
2067   
2371   
2075   
2076   
2202   
1647   
1700   
1626   
1708   
1641   
1657   
1926   
1618   
1610   
2086   
354   
1440   
1211   
900   
1611   
1500   
163   
1625   
2096   
829   
2108   
1149   
701   
157   
1490   
1017   
2107   
1653   
1877   
621   
1519   
1859   
1933   
2119   
2360   
146   
2390   
1372   
2118   
2142   
2130   
2384   
2129   
1668   
1510   
2141   
2267   
2150   
383   
1028   
1031   
479   
1541   
2149   
987   
1659   
462   
520   
1867   
2271   
1281   
708   
2163   
707   
2266   
1680   
534   
2277   
1390   
1598   
538   
2450   
178   
2270   
409   
2153   
2285   
167   
2276
 
    
Let me know if you need any other information regrding same. Thanks. 
On Fri, Oct 07, 2011 at 03:33:39PM -0700, Hardik Sharma wrote:
>After some analysis on my results, I am thinking that the error I am getting is a possible bug.  >>My problem and the analysis is as follows- >I encoded the yuv foreman sequence with 300 slice size, bitrate 512kbps, framerate 30 and GOP size 20 frames (aka keyint 20). I am getting 2488 number of slices for this 300 frame video sequence.       >>I tried to check some values by dropping each and every slice one by one. Some couple of 200 slices are such that out of 2488 slices that if I drop any of those slice, the error goes beyond the gop all over the remaining frame. It is defeating the purpose of keeping gop size and putting key frame in the start of every gop. Also as the error spread from that slice to all the remaining frames, it also drops frame during decoding. I got a drop of around 25-27 frames and thats a lot to drop your PSNR drastically. Please let me know regarding this problem.  >>Please provide a reproduceable testcase.


More information about the ffmpeg-devel mailing list