Timeline for basic hsb skin detection, neon illumination
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 22, 2012 at 17:11 | comment | added | Dima | You can think of the covariance as defining an ellipsoid in 3D. You may be able to visualize it in Matlab, but that is likely to be a lot of work. Alternatively, you can try to look at 2D projections of the ellipsoid, but that would also take some work. | |
| Mar 22, 2012 at 16:16 | vote | accept | nkint | ||
| Mar 22, 2012 at 16:16 | comment | added | nkint | just last thing, i'd like to draw my covariance matrix but i don't know how to best visualize informations of a 3x3 covariance matrix. this means that i don't have really got its graphical meaning. some advice? | |
| Mar 22, 2012 at 16:16 | comment | added | nkint | ok it was a problem about casting from float to double and cast it back (cause threshoal could be done only on float). thanks. it works well. and your explanation is great! | |
| Mar 22, 2012 at 13:31 | comment | added | Dima | @nkint please see edit. | |
| Mar 22, 2012 at 13:31 | history | edited | Dima | CC BY-SA 3.0 |
added 550 characters in body
|
| Mar 22, 2012 at 8:39 | comment | added | nkint | but it is normal to have a covariance matrix with value upper than 1000000 ? what does the mahalanoise distance should be like? number between what range? | |
| Mar 22, 2012 at 0:30 | comment | added | Dima | @nkint, I think I see your problem. The mean will be a vector of floats, and when you subtract it from the pixel values, you will also get floats, some of which will be negative. That means before you do anything, you should store $P$ matrix as floats or doubles. | |
| Mar 21, 2012 at 20:52 | comment | added | nkint | yeah i think i'm doing something wrong because my covariance matrix is: [10913058.00000000, 7046611.50000000, 3290781.50000000; 7046611.50000000, 4811646.00000000, 2225078.00000000; 3290781.50000000, 2225078.00000000, 1387631.87500000] | |
| Mar 21, 2012 at 20:40 | comment | added | nkint | ok great. in 5 lines you made me understand what covariance is. thanks. it works. but i have problem in storing results. if i have pixels from 0-255 what kind of numbers should i have to expect of Mahalanobis distance? if i store them in a 8-uint it does take only a little part of skin, if i store them in a 32-float i get strang white noise | |
| Mar 21, 2012 at 13:11 | history | edited | Dima | CC BY-SA 3.0 |
edited body
|
| Mar 21, 2012 at 12:47 | comment | added | Dima | @nkint, please see the edited answer. | |
| Mar 21, 2012 at 12:46 | history | edited | Dima | CC BY-SA 3.0 |
added 530 characters in body
|
| Mar 20, 2012 at 23:12 | comment | added | nkint | i have problem in understanding how get 3x3 covariance matrix with opencv from an image.. can you give me some reference? | |
| Mar 5, 2012 at 14:40 | history | answered | Dima | CC BY-SA 3.0 |