use spatial error concealment if we dont have a previous frame

Originally committed as revision 1303 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/error_resilience.c

@@ -528,6 +528,8 @@ score_sum+= best_score;
     
 static int is_intra_more_likely(MpegEncContext *s){
     int is_intra_likely, i, j, undamaged_count, skip_amount, mb_x, mb_y;
+    
+    if(s->last_picture.data[0]==NULL) return 1; //no previous frame available -> use spatial prediction
 
     undamaged_count=0;
     for(i=0; i<s->mb_num; i++){