PRIORITY BELIEF PROPAGATION VA FOURIER TEZLASHTIRISH ASOSIDA YUZNI FRONTALIZATSIYA QILISH USULI
Keywords:
yuzni frontalizatsiya qilish, MRF, Belief Propagation, patchasoslangan moslashtirish, Lucas–Kanade, Gabor filtrlari, kompyuterli ko‘rish.Abstract
Maqolada yuzni aniqlash tizimlarining real shartsharoitlardagi samaradorligini cheklab qo‘yadigan pozitsiya muammosini hal etish maqsadida patch-asoslangan MRF–BP (Markov Random Fields – Belief Propagation) frontalizatsiya algoritmining takomillashtirilgan modifikatsiyasi taklif etiladi. Klassik usulga ustuvor xabarlarni rejalashtirish (Priority BP), dinamik yorliqlarni kesish va Fourier domenida tezlashtirilgan og‘irlikli Lucas– Kanade algoritmi integratsiya qilinishi natijasida konvergensiya tezligi 3.6 marta oshirildi. FERET, CMU-PIE va Multi-PIE etalon bazalarda olingan natijalar (mos ravishda 95.5%, 98.8% va 89.4% Rank-1 aniqlik) yondashuvning klassik MRF–BP dan o‘rtacha 3.9% ga, generativ TP-GAN dan 1.2% ga ustun ekanligini ko‘rsatadi.
References
1. Deng J., Guo J., Xue N., Zafeiriou S. ArcFace: Additive angular margin loss for
deep face recognition // Proc. IEEE CVPR. – 2019. – P. 4690–4699.
2. Huang R., Zhang S., Li T., He R. Beyond face rotation: Global and local
perception GAN for photorealistic and identity preserving frontal view
synthesis // Proc. ICCV. – 2017. – P. 2439–2448.
3. Arashloo S. R., Kittler J. Energy normalization for pose-invariant face
recognition based on MRF model image matching // IEEE Trans. PAMI. –
2011. – Vol. 33, No. 6. – P. 1274–1280.
4. Komodakis N., Tziritas G. Image completion using efficient belief propagation
// IEEE Trans. Image Processing. – 2007. – Vol. 16, No. 11. – P. 2649–2661.
5. Phillips P. J. et al. The FERET evaluation methodology for face-recognition
algorithms // IEEE Trans. PAMI. – 2000. – Vol. 22, No. 10. – P. 1090–1104.