Mouth/lip detection based on YCbCr color space, in openCv -
hello i'm trying detect area of mouth or lips. trying use the method described in these papers: paper1,paper2. (could post 2 of papers)
they use method utilize colorspace ycbcr because lips have strong red component , weaker blue rest of face. because of can use color space this:
lipmap = (crcr)((crcr)-n(cr/cb))((crcr)-n*(cr/cb))
n = 0.95*((sum(cr*cr)/m)/(sum(cr/cb)/m))
the problem these sources dont explain how calculation made. know how values of cr , cb not 100% sure in range should [0,1] or [0,255], sources state both. after i'm not sure how correct values n calculation lipmap one. numbers supposed normalized before or after calculation has been made. how correct value cr/cb , normalized value, cant seem 1 right.
i have been trying work hours im doing wrong. values in end far big. not in ranges supposed [0,1] or [0,255].
if has input regarding how solve or tips appreciated.
if trying lips appear on face, might worth while using haar cascade find them instead (i don't know if have tried this, or if suitable project)
if need find settings thresholding yuv, still use haar cascade find mouth, pick pixel within region of interest picked out haar know contain pixel on lip.
this pixel provide yuv value random region on lip, can threshold image include pixels similar pixels yuv value (you can figure out how strict on threshold during testing) , should left pixels of mouth (near enough)
hope have explained enough understand.
good luck
Comments
Post a Comment