1. ºKn
³o¦¸±MÃDªº¤º®e¡A¬O¥HStephen Milborrow ªº STASM v1.5 [1] ¬°®Ö¤ß¥h¥[¥H§ï¨}¡A¦]¬°STASM v1.5 ³oÓµ{¦¡¥Dn¬O°w¹ïµLªí±¡ªº¥¿¦V¤HÁy¥h§ä¥X¯S¼xÂI¡A¦b§PÂ_¤@¨Ç¤ñ¸û¸Ø±iªºªí±¡®É¤@¨Ç¯S¼xÂIªº·Ç½T²v¨Ã¤£¬O«Ü²z·Q¡A³o¦¸±MÃD¥Dn´N¬O§Æ±æ ¯à°÷´£°ª¦³ªí±¡¤HÁyªº¯S¼xÂI·Ç½T²v¡C
ÁÙ§â쥻Command Lineªºµ{¦¡¥[¤W GUI¡A¨Ï¨ä¸û¬°®e©ö³Q¨Ï¥ÎªÌ¾Þ§@¡A¤]§ó®e©öÆ[¬Ýµ²ªG¡C¥t¥~ÁÙ°µ¤F¹ï³sÄò¼v¹³ªº³B²z¡A¹Á¸Õ¥hÁYµu³B²zªº®É¶¡¡C
2. STASM
2.1 STASM Version 1.5
STASM ªº¥þ¦W¬° Stacked Trimmed Active Shape Model ¬O Stephen Milborrow ¦b½×¤å [1] ¤¤©Ò´£¥X¨Óªº¤@Ó¤èªk¥Î¥H§ä¥X¤HÁy¤¤ªº¯S¼xÂI¡C¦b½×¤å¤¤¦³¸Ô²Óªº¤¶²Ð»P»¡©ú¡A¦b³oÃä¨Ã¤£·|¸Ôz¡C
¥H¤Wªººtºâªk¸`¿ý¦Û½×¤å [1] ªº Section 2.3¡A¥Dn´N¬O¨Ï¥Î Face Detector¡]Rowley [2] ©Î Viola Jones [3]¡^ §ä¨ì¤HÁyªº¦ì¸m¤§«á¡A·|®Ú¾Ú Training ¥X¨Óªº Models µ¹¥¦¤@Ó Start Shape¡A¤§«á¦b±q³oÓ Start Shape ¥h°µ Translation, Rotation, Scaling ¨Ï³oÓ Shape ¥h²Å¦X©Ò¿é¤Jªº¤HÁy¡C
Figure 2 ´N¬O ASM Search ¹ê»Ú°õ¦æªº¹Lµ{¡A±q¤@¶}©lªº Shape ©¹¤HÁy¥h°µ¦¬ÀÄ¡C
2.2 STASM Version 1.5 ªºµ²ªG
¦bÆ[¹î¹L STASM ¹ï©ó¤@¨Ç¦³ªí±¡ªº¤HÁyªº°õ¦æµ²ªG¤§«áµo²{¡A¦³®ÉÔ¯S¼xÂIªº»~®tÁÙ¤£¤p¡A¤ñ¸û©úÅ㪺³¡¦ì´N¬O¼L¤ÚªþªñªºÂI¡C
Figure 3 ¬°¤@²Õ Database ¤¤¡u§Ö¼Ö¡vªí±¡ªº¨ä¤¤´X±i¹Ï¡A¥i¥H¬Ý¥X¨Ó¡A¦b²´·ú¡B»ó¤l¸ò½ü¹øªºµ²ªG³£Áٺ⤣¿ù¡A»~®t¤ñ¸û¤jªº´N¥u¦³¦b¼L¤Úªþªñ¡A³o¬O¦]¬° STASM 쥻ªº³]p´N¬Oµ¹¡uµLªí±¡ªº¥¿¦V¤HÁy¡v°µ§PÂ_¥Îªº¡A©Ò¥H¦b§PÂ_ªí±¡¤ñ¸û¸Ø±iªº¤HÁy®É´N·|«Ü®e©ö²£¥Í»~®t¡C
Figure 4 ´N¬O¨S¦³ªí±¡ªº°õ¦æµ²ªG¡A¥i¥H¬Ý¥X¨Ó»~®t¤ñ°_ Figure 3 ¤p¤F¤£¤Ö¡C´î¤Ö¼L¤Úªþªñªº¯S¼xÂIªº»~®t´N¬O³o¦¸±MÃDªº¤@Ó«ÂI¡C
3. §ï¶i¤èªk
3.1 ¼L¨¤
¦b¼L¤Úªþªñªº¯S¼xÂI¤¤¡A§ÚÌ¿ï¾Ü¥ý±q¼L¨¤¶}©l½Õ¾ã¡C¥ý¥H쥻¯S¼xÂIªº®y¼Ð§ä¥X¥]§t¾ãÓ¼L¤Úªº¤@Ó¯x§Î¡]§Q¥Î쥻ªº¨âÓ¼L¨¤¥H¤Î»ó¤l³Ì¤U±ªº¤@ÓÂI¥i¥Hºâ¥X¯x§Îªºªø¸ò¼e¡^¡A¨Ã±N¨äÂà´«¬°¦Ç¶¥¼v¹³¡A¦p Figure 5¡C
¦A¹ï³o±i¹Ï°µ Histogram Stretch¡A§â¼v¹³¤¤ªº¦Ç¶¥È½d³ò©Ô¨ì 0~255¡A¨Ï¨ä¹ï¤ñ§ó¥[©úÅã¡AFigure 6 ¬° Figure 5 °µ Histogram Stretch ªºµ²ªG¡C
¥Ñ Figure 6 ¥i¥H©úÅ㪺¬Ý¥X¨Ó¦b¾ã±i¼v¹³¤¤¡A¼L¨¤¬OÃC¦â³Ì²`ªº³¡¥÷¡A¬°¤F¨Ï¼L¨¤§ó©úÅã¡A§ÚÌ¿ï¾Ü¦b 0~255 ¤§¶¡§ä¥X¤@Ó Threshold È¡A¦Ç¶¥È¤ñThreshold¤pªºÂI³£·|³Q³]¬° 0¡A¦Ó¤ñ Threshold¤jªº³£·|³Q³]¬°255¡A¦p¦¹¥i¥H§â¤@±i¦Ç¶¥¹ÏÂà´«¬°¤@±i¥u¦³¶Â¸ò¥Õ¨âºØÃC¦âªº Binary Image¡C
¦b¹ê°µ®É¡A§Ú̬O®Ú¾Ú Otsu ªººtºâªk [4] ¨Ó§ä¥X Threshold È¡C¸g¹L¹êÅ礧«áµo²{¡A¦pªGª½±µ¥Î쥻ªº Threshold ¨Ó°µ Binary Image ªº¸Ü®ÄªG¨Ã¤£²z·Q¡A¸g¹L½Õ¾ã¡A³Ì«á¿ï¾Ü¥Î쥻 Threshold ªº¤@¥b¨Ó°µ Binary Image ®ÄªG·|¤ñ¸û¦n¡AFigure 7 ´N¬O³oÓºtºâªk³B²z¤§«áªºµ²ªG¡C
¦b Figure 7 ¤¤¡A³Ì¥ª¤Î³Ì¥kÃ䪺ÂI¤À§O´N¬O¨âÃ䪺¼L¨¤¡A¦]¦¹¦b¹³ Figure 7 ³o¼Ëªº Binary Image ¤¤n§ä¨ì¼L¨¤¡A¥un¤À§O±q³Ì¥ªÃä¸ò³Ì¥kÃä¶}©l¥Ñ¤W©¹¤U¤@¦æ¦æ§ä¡A©Ò§ä¨ìªº²Ä¤@ÓÂI´N·|¬O¼L¨¤¡C
3.2 ¼L®B
§ä¨ì¤F¼L¨¤¤§«á¡A±µ¤U¨Ón×¥¿ªº¬O¼L®B¡A¦b STASM ©Òµ¹ªº Model ¤¤¡A¼L®BÁ`¦@¦³16ÓÂI¡A¦ý§Ú̹ê»Ú¥h×¥¿ªº¥u¦³9ÓÂI¡A¨ä¾lªºÂI³£¬O¥Î³o¤EÓÂI¥H¤Î¼L¨¤ªº¨âÓÂIªº¬Û¹ï¦ì¸m¥h¼Ð©w¥X¨Ó¡C
³o¤EÓÂI¤À§O¬O¤W¼L®Bªº¤W½t¡B¤W¼L®Bªº¤U½t¥H¤Î¤U¼L®Bªº¤W½t¦U¤TÓ¡C¦]¬°¼L®BªºÃC¦âÅܤƨS¦³·Q¼L¨¤¨º»ò¦h¡A©Ò¥H¨S¦³¿ìªk«Ü²³æªº±q Binary Image ¤¤¬Ý¥X¨Ó¡C¦b³oÃä§Ú̬O¨Ï¥Î Gradient ¨ÓÆ[¹î¨Ã§ä¥X¼v¹³¤¤ªºÃä½t©Ò¦b¡CFigure 8 ¬°ì¥»¼v¹³¥H¤Î¨ä Gradient ªº¹ï·Ó¹Ï¡C
¦b³oÃä¨Ï¥Îªºpºâ Gradient ¤èªk¡A¦p Figure 9 ©Ò¥Ü¡AX ¤è¦Vªº¤À¶q¬O¥Î«á¤@¦æ°Å«e¤@¦æ¡AY ¤è¦Vªº¬O¥Î«á¤@¦C´î«e¤@¦C¡A³Ì«á¤@¦C»P³Ì«á¤@¦æ³£³]¬°0¡A³Ì«á¦A¨ú¥¤è©Mªº¥¤è®Ú¡C¦b Figure 8 ¤¤¥i¥H¬Ý¥X¨Ó¡AGradient ´N¬O¥Î¨Ó¥YÅãÃä½t¥Îªº¡AÃC¦âÅܤƶV¤jªº¦a¤è¦b Gradient ªº¼v¹³¤¤·|¶V©úÅã¡C
±µ¤U¨Ó§ÚÌ´N¥i¥H¥Î쥻¼L®Bªº®y¼Ð¦b³o±i Gradient Image ¤¤§ä¨ì·sªº®y¼ÐÂI¡C¥H¤W¼L®Bªº¤W½t¬°¨Ò¡A¥i¥H¨D¥X¤TÓÂIªº®y¼Ð¦b Gradient Image ¤¤ªºÈ¡A¦A§â¥¦¥[°_¨Ó¡A´N¬O³o¤TÓÂIªº Gradient ¤§©M¡C¦b¤TÓÂIªº¬Û¹ï¦ì¸m¨S¦³§ïÅܪº±¡ªp¤§¤U¡A¨C¦¸©¹¤W²¾°Ê¤@Ó Pixel ¦Aºâ¤@¦¸ Gradient ªº©M¡A¦b§ä¥X³o¨Ç©M¸Ì±ªº³Ì¤jÈ¡A´N¬O·sªº¤W¼L®B¤W½t¡C³oÓ°j°éªº°õ¦æ¦¸¼Æ¬° Figure 5 ¤¤¯x§Î¤§°ªªº¤K¤À¤§¤@¡C¥H¦P¼Ëªº¤èªk¡A¨Ï¤W¼L®Bªº¤U½t¤]©¹¤W²¾°Ê¡A¦Ó¤U¼L®Bªº¤W½t¬O©¹¤U²¾°Ê¡A¦p¦¹´N¯à¨D±o·sªº®y¼Ð¡C
3.3 ¾ãÅé
×¥¿§¹¼L¤Ú³¡¦ì¤§«á¡A±µ¤U¨Óªº¨BÆJ¬O¹ï©Ò¦³ªºÂI¡]¦@68ÓÂI¡^¦A°µ¤@¦¸×¥¿¡A×¥¿ªº¤èªk¤]¬O°ò©ó Gradient Image¡C
¸ò¼L®Bªº¤èªkÃþ¦ü¡A³£¬O¥h§äªþªñ Gradient Ȥñ¦Û¤v¤jªºÂI§@¬°·s®y¼Ð¡C¦b³oÓ³¡¥÷¡A·j´Mªº½d³ò¡A¬O¥H¨CÓÂI¬°¤¤¤ß¡A´M§ä¦b X¡Ó2¥H¤ÎY¡Ó2 ªº½d³ò¸Ì Gradient Ȥñ¦Û¤v¤jªºÂI¡A¥H¨ä§@¬°·s®y¼Ð¡C
¦b Figure 10 ¤¤¡A¥ªÃ䪺¹Ï¬O쥻ªºµ²ªG¡A¥kÃä¬O×¥¿¤§«áªºµ²ªG¡A¬Ý±o¥X¨Ó®t²§³Ì¦hªº³¡¦ì´N¬O¼L¤Ú¡A¤£¹L¦b³o±i¹Ï¤¤¤W¼L®Bªº³¡¦ìÁÙ¬O¤£°÷·Ç½T¡C
4. ²Îpµ²ªG
Figure 11 ¬°¹ï¤@²Õ205±i¤HÁy¼v¹³ªº³B²zµ²ªG¡A¦]¬°¨C¤@±i¼v¹³³£¦³¤â°Ê¼Ð©wªº¯S¼xÂI¡A©Ò¥H´N¥H¦¹¬°°ò·Ç¨Ó¤ñ¸û×¥¿«e«áªº®t²§¡C
¤â°Ê¼Ð©wªº¯S¼xÂI¦³83Ó¡A¦ÓSTASM ªº¿é¥XÂI¦³68Ó¡A¸g¹L¤ñ¸û¤§«á¡A§ÚÌ¿ï¾Ü¤F46ÓÂI¬°¥¦Ìªº¥æ¶°¡A¬O¥H¤H¤u§PÂ_ªº¤è¦¡§ä¥X³o¨Ç¬Û¹ï¦ì¸m®t¤£¦hªº¯S¼xÂI¡A¤]´N¬O»¡¨S¦³¿ï¨ìªºÂI¬O¦]¬°¨S¦³¬Û¹ïÀ³ªºÂI¡C
Figure 11.1 ¬° newPointMean ¡V oriPointMean¡CnewPointMean ªº·N¸q¬°·s±o¨ìªº®y¼Ð»P¤â°Ê¼Ð©wªº®y¼Ð¤§¶ZÂ÷¡A¦Ó oriPointMean ¬°ì¥» STASM ªº®y¼Ð»P¤â°Ê¼Ð©wªº®y¼Ð¤§¶ZÂ÷¡A¨âªÌ¬Û´î´N¬O×¥¿«e«áªº®t¶Z¡C¹Ï¤¤ªº¾î¶b¤À§O¬°²Ä1~²Ä46ÓÂI¡A¬Û¹ïÀ³ªºÁa¶b§Y¬°¸ÓÂIªº¥§¡×¥¿¶q¡A³æ¦ì¬° Pixel¡A¦pªG¤j©ó¹s´Nªí¥Ü×¥¿«á¶ZÂ÷ÅÜ»·¡A¦Ó¤p©ó¹s´Nªí¥Ü×¥¿«áªº¶ZÂ÷Åܪñ¡A³æ¦ì¬O Pixel¡C¥HFigure 11.1 ªº46ÓȦӨ¥¡A¹ï©ó¨CÓÂI¥§¡¤U¨Ó¬O×¥¿¡]Åܪñ¡^¤F 0.02 Ó Pixel¡C
Figure 11.2 ¬° newPicMean ¡V oriPicMean¡CnewPicMean ªº·N¸q¬°×¥¿¤§«á¡A¹ï©ó¨C±i¼v¹³¦Ó¨¥¡A·s±o¨ìªº®y¼Ð»P¤â°Ê¼Ð©wªº®y¼Ð¤§¥§¡¶ZÂ÷¡A¦Ó oriPicMean ¥Nªíªº«h¬O×¥¿¤§«eªºÈ¡C¹Ï¤¤¾î¶b©Ò¥Nªíªº´N¬O²Ä1~²Ä205¼v¹³¡A¬Û¹ïÀ³ªºÁa¶b§Y¬°¸Ó¼v¹³ªº¥§¡×¥¿¶q¡A³æ¦ì¸ò Figure 11.1 ¤@¼Ë³£¬O Pixel¡C¦]¬°¬O¦P¤@¥÷¸ê®Æ°µªº²Îp¡A©Ò¥H¥§¡¤U¨Ó¤]¬O Ÿì0.02¡C
5. ¥[³t³sÄò¼v¹³ªº³B²z
³o³¡¥÷n§ï¶iªº¬O¹ï©ó³sÄò¼v¹³ªº³B²z³t«×¡A¥iÀ³¥Î©óªí±¡°Ê§@¼v¹³ªº°lÂÜ¡C쥻ªºµ{¦¡¬O§â¨C¤@±i¿é¤Jªº¼v¹³³£µø¬°¿W¥ßªº¡A¤À§O¹ï¥¦Ì¥h°µ³B²z¡A©¼¦¹¤¬¤£¬Û¤z¡C¦pªG²{¦bªº¿é¤J¬O³sÄòªº¦n´X±i¼v¹³¡A¤]´N¬O³sÄòªº Frame ªº¸Ü¡A¹ï©ó¨C¤@±i¼v¹³³£·|°µ¤@¦¸ Figure 2 ªº¨BÆJ¡A±q³Ì¤@¶}©lªº Shape¥h°µ¦¬ÀÄ¡C
¦]¬°¬O³sÄòªº¼v¹³¡A©Ò¥H«e«á¨â±i Frame ªº®t²§¤£·|¤Ó¤j¡C°ò©ó³oÓ·Qªk¡A§ÚÌ¥t¥~´£¥X°w¹ï³sÄò¼v¹³ªº³B²z¤è¦¡¡C¤]¬O´N¥u¦³²Ä¤@Ó Frame ·|±q¤@¶}©lªº Shape ¶}©l°µ¦¬ÀÄ¡A¦b²Ä¤@±i Frame ¤§«áªº¼v¹³¡A³£¬O¥Î«e¤@±i¼v¹³¦¬ÀĤ§«áªºµ²ªG§@¬°¨ä Start Shape¡C¥Î³oÓ¤èªk¨Ó¬Ù²¤¤HÁy°»´ú¥H¤Î«e±¦¬ÀĪº¨BÆJ¡A¤w¹F¨ì§ó§Öªº³B²z³t«×¡C
¦b³oÓ¤èªk¤¤¡A§Ṳ́]¥un¹ï²Ä¤@±i Frame °µ Face Detection¡A¤§«áªº Frame ³£¤£»Ýn¦A°µ¦Ó¬Oª½±µ¥Î²Ä¤@±i Frame ªºµ²ªG¡C³o¼Ë³B²zªº®ÄªG¡A¦b¤HÁy¨S¦³¤Ó¤jªº¦ì²¾®É¬O¬Û·í¤£¿ùªº¡C
Figure 12 ¬O¹ï³sÄòªº12±i¼v¹³³B²z©Ò¤À§O¯Ó¶Oªº®É¶¡¡A²`¦â( Ind¡A¥Nªíªº¬OIndependent )ªº¬O쥻ªº¤èªk¡A§â³o12±i¼v¹³µø¬°¿W¥ßªº¡A¤À§O¥h°µ³B²z¡C²L¦â( Seq¡A¥Nªíªº¬OSequential )ªº¬O¥Î¤Wz¤èªkªº°õ¦æµ²ªG¡C
¤£¬Ý²Ä¤@±i¼v¹³ªº¸Ü¡A¨ä¥Lªº¼v¹³¤j·§³£ÁYµu¤F30%~40% ªº®É¶¡¡AÁ`Åé¦Ó¨¥ÁYµu¤F29%ªº®É¶¡¡C²Ä¤@±i¼v¹³¦]¬°¨âºØ¤èªk©Ò°µªº³£®t¤£¦h¡A©Ò¥H®É¶¡¨S¦³¤Ó¤jªº®t¶Z¡C
¦Ü©ó¬°¤°»ò쥻ªº¤èªk¤]·|¨Ï±o²Ä¤GÓ¥H«áªº³B²z®É¶¡¤]ÁYµu¡A§Ú̲q´ú¥i¯à¬O¦]¬°²Ä¤GÓ¥H«áª½±µ¨Ï¥Î²Ä¤@Óªº°O¾ÐÅéªÅ¶¡¡A¤£¥Î¦Aªá¤Ó¦h®É¶¡°t¸m¡A©Ò¥H¨Ï¦s¨úªº®É¶¡ÁYµu¡C
6. ¨Ï¥ÎªÌ¹Ï§Î¤¶±
¦¹µ{¦¡ªº®Ö¤ß¬O STASM v1.5¡A¦Ó¹Ï§Î¤¶±¬Oקï¦Û The Yovav (Horror) PictureShow Version 1.00 [5]¡C¦¹¥~ÁÙ¥[¤W¤@¨Ç¥i¥HÅý¨Ï¥ÎªÌ¿ï¾Üªº¿ï¶µ¡A¦p Figure 13 ªº¥k¹Ï©Ò¥Ü¡A¨Ï¥ÎªÌ¥i¥H¿ï¾Ü¿é¤Jªº¼v¹³ºØÃþ¡B¤HÁy°»´ú¾¹¡B¿é¥X°£¤F¼v¹³¤§¥~ÁÙn¤£n¿é¥X¯S¼xÂIªº®y¼Ð¡C
7. µ²½×
¦]¬°¥Ø«e¥u¦³¾Ç¹L²³æªº¼v¹³³B²z§Þ¥©¡A±q²Îp¤Wªºµ²ªG¨Ó»¡ÁöµM¬O¦³§ï¶i¡A¦ý¬O§ï¶iªº´T«×¨Ã¨S¦³«Ü¤j¡A¦pªG·QnÅý§PÂ_¯S¼xÂIªº·Ç½T«×¦³§ó©úÅ㪺´£°ª¡A´Nn¹ïµ{¦¡¶i¦æ¸û¤j³W¼Ò¥B¸û²`¤Jªº×§ï¡A¹ï©ó²{¶¥¬qªº§Ú¦Ó¨¥¡A¥Ø«eÁÙ¨S¦³¯à¤O¸Ñ¨M¦¹°ÝÃD¡A³oÓ°ÝÃDnµ¥¨ì¥¼¨Ó°µ§ó²`¤Jªº¬ã¨s¤§«á¤~¦³¥i¯à¸Ñ¨M¡C
8. °Ñ¦Ò¸ê®Æ