1 00:00:00,000 --> 00:00:03,000 We will now talk about correspondence in stereo. 2 00:00:03,000 --> 00:00:06,000 You might remember our dynamic programming approach for 3 00:00:06,000 --> 00:00:09,000 resolving correspondence along an entire scan line. 4 00:00:09,000 --> 00:00:15,000 So I'll give you another scan line. This is the left scan line--red, red, blue, blue, blue, red. 5 00:00:15,000 --> 00:00:19,000 Then the right scan line we get to see the following. 6 00:00:19,000 --> 00:00:22,000 Obviously there is a shift going on. 7 00:00:22,000 --> 00:00:27,000 I'd like to ask you where this little pixel over here will go into the lead association. 8 00:00:27,000 --> 00:00:31,000 It can go into any of those pixels over here, so please check exactly one of those boxes. 9 00:00:31,000 --> 00:00:34,000 Let's assume the cost for a bad match, 10 00:00:34,000 --> 00:00:37,000 when we match 2 colors that don't correspond, is 20. 11 00:00:37,000 --> 00:00:41,000 The cost of an assumed occlusion or a disocclusion is 10. 12 00:00:41,000 --> 00:00:43,000 Try to find the optimal alignment, 13 00:00:43,000 --> 00:00:49,000 and then tell me where in the right scan line this 1 pixel corresponds to. 14 00:00:49,000 --> 00:00:52,000 Check the exact box to which it corresponds. 15 00:00:52,000 --> 00:00:54,000 Here is a second question I'd like to ask you. 16 00:00:54,000 --> 00:00:58,000 What if we changed the cost of occlusion to 100? 17 00:00:58,000 --> 00:01:01,000 Please answer the exact same question--where does the B over here go-- 18 00:01:01,000 --> 99:59:59,999 under this different cost model.