TẠO DỰNG ẢNH HIỂN VI VỚI ĐỘ PHÂN GIẢI SIÊU CAO PHỤC VỤ CHO CÁC ỨNG DỤNG KHOA HỌC

DAI HQC QUOC GIA HA NOI Tao dirng anh hien vi vol phan giai sieu cao phuc vu cho cac u*ng dung khoa hoc Ma so: QC.09.07 Chu nhiem de tai: Le Vu Ha DAI HOr OLIOC C^IA HA NOI TRUNG 1AM \HON^ ilN IHU VIEN 000600000"^^ HaN6i-2010 MUC LUC • • Bang giai thich cac chu* viet tat Danh sach nhirng ngvoi tham gia thyc hien de tai 3 Danh muc cac bang so li^u Danh myc cac hinh ve Tom tdt cdc ket qua nghien cmi chinh cua de tai BAO CAO T N G K £ T 6.1 Dat van de 6.2 Tong quan cac van de nghien cihi 6.3 Muc tieu va Not dung Nghien cihi 6.3 L Xay dung thuat toan tao dung anh SRtii cac anh LR 6.3.2 Ket qua thuc nghiem va ket luan 15 Ket qua so sanh cac phuong phap n^i suy duo'c the hien cac bang sau day, cac gia tri bang la PSNR ciia anh HR duac tai tao lai tu cac anh LR mo phong, so vol anh HR ban dau 19 Chung ta co the thay rang hau het cac tnrong hpp phuong phap KRI deu cho ket qua tot nhat 20 6.4 Thoi gian, dia diem va phuong phap nghien ciru 20 6.5 Ket qua nghien cuu 20 6.5.1 Cac cong bo lien quan den ket qua cua de tai 20 6.5.2 Ket qua dao tao cua De tai 20 6.5.3 Ket qua ung dung cua De tai 21 6.6 Thao luan 21 6.7 Ket luan va kien nghj 21 6.8 Tai lieu tham khao 21 PHU LUC (cac giay t^ gui kem) 23 Bang giai thich cac chu viet tat SR (Super Resolution) HR (High Resolution) LR (Low Resolution) PSF (Point Spread Function) CCD (Charge Coupled Device) MAP (Maximum A Posteriori) MSE (Mean Squared Error) Do phan giai sieu cao Do phan giai cao Do phan giai thap Ham nhoe (ham truyen dat cua thiet bi tao anh) Thi^t bi tich dien kep (kilu cam bien dung thiet bi chup anh so) Hau nghiem cue dai (tieu chuan diing tinh toan toi uu/uac luong) Sai so binh phuomg trung binh Danh sach nhung nguoi tham gia thyc hien de tai Hoc vj Caquan cong tac Bo mon Xu ly thong tin Tien sT Bp mon Vi co dien tu Tien sT Nguyen Thang Long va Vi he thong Thac sT Bp mon Xu ly thong tin Nguyen Hong Thinh Tran Hoang Tung Thac sT Bp mon Xu ly thong tin Nguyen Viet DQng Cunhan Bp mon Xu ly thong tin Nguyen Thuy Hoang Phong Cunhan Bp mon Xu ly thong tin TT Ho va ten LeVuHa Nhiem vu Chu nhiem de tai Thanh vien Thanh vien Thanh vien Thanh vien Thanh vien Danh muc cac bang so lieu Bang 1: PSNR ciia cac anh HR tai tao tu chuoi anh Lena Bang 2: PSNR cua cac anh HR tai tao tir chuoi anh Vase Bang 3: PSNR cua cac anh HR tai tao tir chuoi anh Apaca 19 19 20 Danh muc cac hinh ve Hinh 1: Mo hinh qua trinh tao anh Hinh 2: Lien he giua anh dp phan giai cao va cac anh dp phan giai thap Hinh 3: Vi du ve lay mau tren tin hieu Hinh 4: Qua trinh tao dung anh SR tir tap anh LR 10 Hinh 5: Trai: mpt anh LR chuoi anh Apaca, Phai: anh SR (dp phan giai tang 5x5 l^in) 15 Hinh 6: Trai: mpt anh LR chuoi anh Emily Phai: anh SR (dp phan giai tang 5x5 1^) 15 Hinh 7: Trai: mpt anh LR chuoi anh Text Phai: anh SR (dp phan giai tang 5x5 ikn) 16 Hmh 8: Trai: mpt anh LR chuoi anh Vase Phai: anh SR (dp phan giai tang 5x5 \kn) •; .116 Hinh 9: Trai: mpt anh LR chuoi anh Lena Phai: anh SR (dp phan giai tang 5x5 l4n) 17 Hinh 10: Hai anh mpt chu6i anh chup bang thiet bi hien vi Hinh 11: Cac anh SR hien vi duac tao dung tir chudi anh LR 17 18 Tom tat cac ket qua nghien cuu chinh ciia de tai Ket qua ve khoa hgc (nhung dong gop cua de tai, cac cong trinh khoa hpc da cong bo) De tai da dat duac nhung ket qua chinh ve khoa hoc sau day: • Nghien cuu cac phuang phap tao anh voi dp phan giai sieu cao tir nhieu anh a dp phan giai binh thuong, ap dung cho anh hien vi quang hpc • Nghien cuu so sanh cac phuang phap noi suy khac su dung qua trinh tao dung anh sieu phan giai • Danh gia dp chinh xac ciia cac phucmg phap tao dung anh dugc ap dung Mot phan cac ket qua tren da duac cong bo bao cao khoa hoc sau day: Hong-Thinh Nguyen, Viet-Dung Nguyen, Ha Vu Le, "Super-Resolution Image Construction from High-Speed Camera Sequences", Proceedings of the International Conference on Advanced Technologies for Communications., Ho Chi Mmh City, October 2010 Ket qua phuc vu thuc te (cac san pham cong nghe, kha nang ap dung thuc te) • Chuang trinh cai dat tren Matlab ap dung dk tao anh dp phan giai cao tir chudi cac anh dp phan giai thap hoac video dp phan giai thap Ket qud dao tao (so lupmg sinh vien, so lugng hpc vien cao hpc, nghien cuu sinh tham gia thuc hien lam viec de tai, so khoa luan, luan van da hoan va bao ve) • 02 sinh vien dai hpc va 02 hpc vien cao hpc da tham gia lam viec va nghien cuu khuon kh6 dk tai • 01 khoa luan tot nghiep dai hpc va 01 luan van cao hpc theo huong cua de tai da bao ve cong • 01 cong trinh NCKH cua sinh vien tham gia de tai dat giai nhat c4p truang va giai khuyen khich cap Bp GD-DT Ket qud nang cao tiem lire khoa hpc (nang cao trinh dp can bp va trang thiet bi hgc phhn mem da xay dung dugc giao nop dua vao su dung tai don vi) • De tai da gop phan phat trien mot bp thu vien phan mem xir ly anh hieu qua cho phep tiep tuc dugc khai thac tiep cho phuc vu nghien cuu va dao tao • De tai da tang cuang kien thuc cho cac can bp cua bp mon ve xu ly tin hieu, xu ly anh va video BAO CAO TONG KET 6.1 Dat van de Dp phan giai ciia hinh anh la mpt chi tieu chat lugng quan trpng ciia cac thilt bi chup anh so, dugc bang mot dai lugng ty le nghich voi kich thuac moi diem anh tren be mat cua cam bien hinh anh dimg thiet bi chup anh so Voi viec cac thiet bi hiln thi hinh anh (TV, may in, man hinh may tinh ) co dp phan giai cang cao, dp phan giai cua cac thilt bi chup anh ciing c ^ phai dugc nang cao mpt each tuong irng de tan dung dugc nang luc ma cac cong nghe hiln thi hinh anh mang lai Tuy nhien, neu luon thay the cac thiet bi chup anh dang dugc su dung rit rpng rai rat nhieu Imh vuc, tir cac lihh vuc khoa hpc — cong nghe cho toi cac Imh vuc giai tri, bang cac thiet bi co dp phan giai cao han thi su lang phi se rat Ion vi loai thiet bi co gia rat cao, nhat la vai cac thiet bi chup anh chuyen dung su dung cong nghiep ciing nhu nghien cuu khoa hpc Do do, ton tai mpt nhu cau thuc te cho cac cong cu xu ly anh cho phep tao dung nhung buc anh co chat lugng tuang ducmg san pham cua cac thiet bi chup anh co dp phan giai cao tir anh dugc chup bdi cac thiet bi co dp phan giai thap han Tao anh sieu phan giai tir cac anh co dp phan giai thap la van de da dugc nghien cuu tir nhieu nam va van tiep tuc la chu de co tinh thai sir cong dong xir ly anh Thuat ngu "sieu phan giai" dugc hieu vai nghia dp phan giai cua anh dugc tao cao han dp phan giai thuc cua thiet bi chup anh Hang tram phuang phap tao anh sieu phan giai da dugc de xuat, voi hang chuc phuang phap da tra kinh dien (Ur&Gross, 1992; Tekalp et al, 1992; Komatsu et al, 1993; Hardie et al, 1998; Baker&Kanade, 1999; Elad&Feuer , 1999; Nguyen&Milanfar, 2000; Altunbasak&Patti , 2000; Joshi&Chaudhuri, 2002; Protter et al, 2009, ) Rieng nam 2010 so lugng bai bao lien quan tai van de cong bo tren hai tap chi CO uy tin nhat Imh vuc xur ly anh la IEEE TPAMI va IEEE TIP cung rat nhieu (Yang et al, 2010; Robinson et al, 2010; Gajjar&Joshi, 2010; Kim&Kwon, 2010; Xiong et al, 2010; Belekos et al, 2010; Yuan et al, 2010; van Eekeren et al, 2010; Mallat&Yu, 2010), The nhung thuc te cac ky thuat tao anh sieu phan giai, nhat la cac phuang phap tir nhieu anh, van rat it dugc ung dung Ly chinh la boi vi cac phuang phap da dugc de xuat mac du co ca sa ly thuyet virng chac va cac ket qua thuc nghiem tot van that bai ung dung dieu kien thuc, Mpt so nguyen nhan sau day co thi ly giai dieu do: • Phan Idn cac phuang phap tao anh sieu phan giai tir nhilu anh doi hoi phai xac dinh dugc cac dich chuyen xay giua cac anh vdi dp chinh xac cao (muc dudi diem — subpixel), dd la van de rat kho cac chuyin dpng xudt hien chu§i anh rdt phuc tap hoac anh bi nhieu • Trong dieu kien thirc cac anh thu dugc tir video camera co dp phan giai thdp thudng bi nhoe nhieu ly ket hgp nhu camera co tieu cu co dinh, anh hudng cua rung dpng moi trudng, va chuyen dpng cua cac doi tugng nen g ^ nhu khong thi xac dinh chinh xac ham nhoe cua qua trinh chup anh, hiu hit cac phuang phap tao anh da phan giai bo qua anh hudng cua anh nhoe hoac gia thilt ham nhoe ciia he thong cd the xac dinh duac Muc tieu cua chiing toi de tai khong phai la de xuat mpt phuang phap tao anh da phan giai mdi ma la tim mpt linh vuc ung dung thuc tl d cac phuang phap da cd cd thi thuc su phat huy tac dung Vdi muc tieu nhu vay, chiing toi da chpn nghien cuu kha nang ung dung ky thuat tao anh sieu phan giai tir nhieu anh cho anh hiln vi quang hpc (optical microscopic images) vi nhirng ly sau day: • Mac du cac thiet bi hien vi quang hpc thudng co dp phong dai rat Idn va six dung cac thilt bi chup anh cd dp phan giai cao, nhu cau co dugc hinh anh vdi dp phan giai cao hon niia luon luon dugc mong muon, nhat la cac Itnh vuc nghien cuu ve cac cau tnic cd kich co d muc nano • Hinh anh chup bang thiet bi hien vi quang hpc thudng cd chat lugng rat cao dieu kien chup anh thudng dugc kiem scat tot: dilu kien chieu sang tot, it nhieu, it nhoe Trong trudng hgp can thiet viec dac ham nhoe ciia qua trinh chup anh ciing cd the thuc hien dugc kha dl dang • Viec tao chuoi anh vdi cac dich chuyen nho giiia cac anh co the dugc kiem soat de dich chuyen giira cac anh kha don gian va de dang tinh toan/do dac dugc vdi dp chinh xac cao Ky thuat tao anh sieu phan giai neu ap dung cong cho cac anh hien vi se la mpt cong cu hiru dung phuc vu cho cac phong ihi nghiem nghien cuu va phan tich cac lihh vuc nhu vat lieu, y-sinh hpc va y te 6.2 Tong quan cac van de nghien cuu Cd the chia cac phucmg phap tao anh sieu phan giai hai nhdm: cac phuang phap thuc hien tir nhat mot anh, thudng gpi la noi suy anh, va cac phuang phap thuc hien tir nhieu anh U'u diem ciia cac phuang phap noi suy anh la chiing cd dp phuc tap tinh toan thap hon so vdi cac phuong phap thuc hien tir nhieu anh Tuy nhien, nhieu chi tiet bi mat tren anh cd dp phan giai thap hien tugng aliasmg se khdng the khdi phuc lai dugc neu chi sir dung mpt anh, hoac ngugc lai, cac phuang phap npi suy hay tao cac chi tiet khong cd thuc (artifacts), vi vay kho cd thi tao anh sieu phan giai vdi dp chinh xac cao bang cac phucmg phap npi suy don thuin Trong do, anh hiln vi sir dung nghien cuu hay phan tich can phai cd dp chinh xac cao vi chiing dugc diing khong chi de quan sat ma cho cac phep dac kich thudc cac cau tnic Do dd, cac phuang phap tao anh sieu phan giai tir nhieu anh se phii hgp hon ap dung cho anh hiln vi vi cac phuang phap su dung dugc nhieu thong tin chi tiet hon tap hgp tu nhieu anh de tong hgp buc anh cd dp phan giai cao Qua trinh tao anh sieu phan giai tir nhieu anh thudng bao gom cac budc nhu sau: Uac lugng sir dich chuyen giua cac anh nham muc dich bii cac dich chuyin dd dl dang ky cac anh dau vao (gpi la cac anh co dp phan giai thap - LR) len mpt ludi dilm cd dp phan giai cao hon Vi vay cac dich chuyen ciia moi dilm giiia hai anh lien tiep can dugc xac dinh vdi dp chinh xac tuong duong vdi ludi dilm cd dp phan giai cao (HR), nghia la vdi dp chinh xac d muc dudi kich thudc dilm cua cac anh ban dau, Vdi cac chudi anh co chua nhieu chuyin dpng phuc tap, dd la mpt nhiem vu rat khd khan Tuy nhien, vdi cac he thdng chup anh tinh vi nhu cac thilt bi hiln vi quang hpc, viec chii dpng dieu khien de cac chuyen dich giiia cac anh rat nho va don gian la kha de dang, nen viec xac dinh chinh xac cac chuyin dich co thi thuc hien vdi nhiing phuong phap dcm gian Dien dir lieu tir cac anh LR (sau da dugc bii dich chuyen) vao ludi HR Ndi suy cac dilm cdn thieu ludi HK Viec phai dimg den ndi suy la khdng thi tranh khdi chudi anh LR chiing ta cd khdng dii de dien vao th ca cac dilm ludi HR Cd rat nhilu phuang phap ndi suy anh cd thi ap dung dugc, tir cac phuong phap ndi suy toan hpc thudng dimg, ndi suy tin hieu bang ham sine, cac phuang phap thdng ke, den cac phuong phap ndi suy su dung cac dac tnmg hinh hpc ciia anh Dilu quan trpng d day la lira chpn dugc phuang phap ndi suy thich hgp, dam bao tinh chinh xac cao ciia cac chi tiet dugc ndi suy Khu nhoe (deblurring): mpt sd v4n de xay qua trinh chup anh cd the gay hien tugng nhoe lam md di cac chi tiet tren anh, nhu nhoe dat sai tieu cu hay nhoe rung dpng cua mdi trudng anh hudng tdi thilt bi Vdi cac he thdng chup anh hien vi day thuc khdng phai la van de Idn vi: a) khoang each chup thudng la cd dinh nen cd the chi can hieu chinh he thdng mpt lan de cd dp net cao nhat cho ca chudi anh, va b) cac he thdng chup anh hien vi thudng dugc thiet ke cd kem theo he chdng rung de han che anh hudng tir rung dpng mdi trudng Ngoai ra, neu can xac dinh chinh xac ham nhoe cua qua trinh chup anh de thuc hien viec khu nhoe, vdi cac he thdng hien vi ham nhoe thudng dugc mpt each kha chinh xac (do ham nhoe ciia he thdng chup anh chinh la dap ung xung cua he thdng, nen vdi he thdng chup anh hien vi ngudi ta thudng ham nhoe bang each chup mpt diem sang rat nho tren mat phang dat vat mau ciia kinh hien vi, tao bdi lazer, dieu kien hoan toan khdng cd anh sang) Khi cd the ap dung nhieu phucmg phap khir nhoe vdi ham nhoe da biet Cac phuong phap hien co quan tam chu yeu tdi viec cai thien chat lugng hinh anh theo cam quan ciia ngudi hon la tdi dp chinh xac ciia hinh anh, vi vay khdng dam bao viec nang cao chat lugng hinh anh cung se giiip nang cao dp chinh xac cua cac phep phan tich va dac tren anh Cac de tai nghien cuu ve tao anh sieu phan giai vdi ddi tupmg chinh la anh hien vi quang hpc ciing rat it 6.3 Muc tieu va Noi dung Nghien cuu Muc tieu ciia de tai la nghien cuu cac giai thuat tao dung anh sieu phan giai tir nhilu anh nhim dua mpt giai phap ung dung cho anh hien vi quang hpc Chiing tdi chi tap tnmg vao cac giai thuat udc lugng dich chuyen giiia cac anh va cac giai thuat ndi suy anh nhiing ly da dugc trinh bay d phan tren Cac ndi dung nghien ciru dugc trinh bay cu thi cac muc dudi day 6.3,1, Xay dimg thuat todn tao dung dnh SR tir cdc dnh LR Mo hinh bdi todn Ban chit ciia qua trinh chup anh la qua trmh lay mau khdng gian lien tuc cam biin ciia thiet bi anh ddng vai trd nhu mpt bp lay mau tin hieu hai chieu mat p h ^ g anh Anh thu dugc se la ma tran cac gia tri thi hien muc sang toi thu dugc tir cam biin anh tiiy vao dp xa gan, mau sac ciia vat the Ve nguyen tac, tan sd lay mau cang Idn (nghia la kich thudc diem anh tren cam bien hinh anh cang nhd), anh thu dugc se cang ro net; va tan sd lay mau dii Idn (theo tieu chuan Nyquist), anh thu dugc se the hien day du thdng tin ciia khdng gian lien tuc va mat thudng se khdng phan biet dugc su rdi rac cua thdng tm Ta dinh nghia anh la anh cd phan gidi cao ly tuang Tuy nhien, anh thuc ma chiing ta thu dugc qua qua trmh chup anh thudng cd dp phan giai thap hon dp phan giai ly tudng dd Ngudi ta thi hien dilu bSng mot su gidm toe lay mau (dovm-sampling), thudng gpi tat la giam tdc, nghia la giam dp phan giai cua anh Ngoai ra, anh chup dugc cdn la ket qua cua viec biin doi hinh anh ly tudng cua khung canh thuc bdi qua trinh chup anh, dai dien bdi dap ung xung ciia qua trinh nay, thudng dugc gpi la ham nhoe (point spread function) Md hinh qua trinh tao anh dugc minh hpa hinh dudi day: ANH O O PHAN CIAI THAP Hinh 1: Mo hinh qud trinh tao dnh De cd the xay dung thuat toan tao dung anh sieu phan giai (SR), ngudi ta cd gang md hinh hoa mdi lien he giiia anh cd dp phan giai cao (HR) ly tudng va anh cd dp phan giai thip (LR), va mdi lien he giua cac anh LR vdi Neu md hinh dugc xay dung la chinh xac, b ^ g each dao ngugc lai qua trinh tir anh HR->LR, chiing ta hoan toan cd thi tao dung lai anh HR, tuc la anh SR, tir cac anh LR Gpi {¥*} la tap cac anh LR, chup bdi cung mpt may anh, cimg dieu kien, nhung cd su khac sir dich chuyen ciia camera hay ddi tugng dugc chup qua trinh chup anh Gpi X la anh dp phan giai cao ly tudng ma chiing ta mong mudn co dugc Anh LR thu dugc cd thi coi nhu ket qua ciia qua trinh chup anh bdi he thdng cd ham nhoe h va sau giam tdc tir anh dp phan giai cao ly tudng Qua trinh chup anh cdn cd thi bi anh hudng ciia nhieu tir mdi trudng, Neu ky hieu D va H lan lugt la toan tir giam tdc va nhan chap vdi h, va n dai dien cho phan nhieu, chiing ta cd: Y^ = DHXt+n, d X* la mpt phien ban bi dich chuyen (tinh tien va xoay) ciia X Gpi F^ la toan tii dich chuyen tuong ung, chiing ta cd dugc md hinh toan hpc ciia qua trinh tao anh dugc bieu dien bang phuong trinh dudi day: Yk = DHXi+n = DHF;tX+n NH)£U ANH D O PWAN CIAI THAP YK GIAMTOC ^ ^^it^S/-^*• W A N ClAl CAO NHUNC 81 N H E i i - i M ^ _ „ *& Hinh 2: Lien he giira dnh phan gidi cao vd cdc dnh phan gidi thap Ca sa cua thuat todn Gia sir rang sir dich chuyen giua cac hinh anh la hoan toan ngau nhien Cau hoi dat la dieu kien nao chiing ta cd the ket hgrp cac anh LR de thu dugc anh cd dp phan giai cao hon De don gian chiing ta cd the xem xet viec lay mau tin hieu mot chieu /(/.) : Tin liicu lien luc 1/0 C*^) : La\ m5u tai (-0 y i (•"•) : L5y mau tai t-tl M+*iM+*i JV-liV-1 N N + ^1 Hinh 3: Vi du ve lay mau tren tin hieu Tin hieu rdi rac dugc lay mau (sampled) tir tin hieu lien tuc Ta cd the dimg khai niem dp phan giai cua tin hieu dl chi mat dp mau tin hieu hay tan sd lay mau, dai lugng ty le nghich vdi chu ky lay mau (khoang each giira hai mau) Ddi vdi anh sd, mpt chu ky liy mlu (hai chieu), tuong duong vdi kich thudc mpt diem anh 10 TABLE m RESULTS ON SIMULATED SEQUENCE #3 (VASE) tynmtili daru M I nnnai ( # of frames 10 15 20 Fig Linear 28.8 28.78 30.87 28.59 34.5 28.78 KRI 29.2 29.92 31.02 28.63 34.78 29.92 Scattered 27.87 29.31 29.81 28.84 32.71 29.31 Warping data from multiple frames into a HR grid obtained from simulated sequence with added noise For all sequences, we could see using KRI for missing pixel interpolation always achieves the best performance Another observation from these results is that noise could severely degrade the performance of super-resolution methods TABLE RESULTS ON SIMULATED SEQUENCE #1 (LENA) # of frames 10 15 20 Linear 29.2 29.18 31.79 29.18 36.15 30.27 KRI 30.24 30.44 33.24 31.5 36.56 31.58 Scattered 28.3 28.98 30.32 30.58 35.68 31.11 Fig An LR frame from a real-world high-speed camera sequence noise and about 20dB for simulated sequences with added 20dB noise), but the visual quality of constructed images is comparable to those resulted from optical flow-based methods, as shown in Fig.9 A possible explanation is that NLM method tends to smooth irregularities in images TABLE n RESULTS ON SIMULATED SEQUENCE #2 (BARBARA) # of frames 10 15 20 Linear 27.1 24.78 31.43 27.01 35.23 29.79 KRI 29.08 26.28 35.62 27.03 38.01 31.23 Scattered 28.53 24.37 31.59 25.6 33.4 26.94 Fig.7 is an original frame from a real-worid high frame rate sequence we used in our experiments, and the super-resolution image constructed from that sequence is shown if Fig.8 For the NLM-based super-resolution method, the performance is quite low in term of PSNR (about 25dB for simulated sequences without Fig Super-resolution image of a moving car constructed from a real-world sequence This work is partly supported by the project QC.09.07 of the Vietnam National University Hanoi, and by an internship for Ms Hong-Thinh Nguyen provided by lEF REFERENCES Fig Super-resolution image constructed from seven frames of a real-world sequence by using NLM-based method IV CONCLUSIONS AND FUTURE W O R K Super-resolution construction for video containing local moving objects is always a difficult task Using high-speed camera sequences, with can be obtain dense-grid of pixels but also need extremely strong method to decrease time processing for real time purpose However, in this paper, we just evaluate the ability of using high speed camera for moving object image super resolution Methods employing optical flow-based motion estimation or NLM (no motion estimation needed) are good candidates However, from our experiments with these methods, we have realized that the key for improving quality of constructed images lies with the selection of image interpolation techniques State-of-the-art methods in image interpolation are the ones making use of geometric regularities like [16] Our findings are quite consistent with this trend since the use of KRI always yields the best results The disadvantage of [16] and some other image interpolation methods is they are developed originally for uniform-grid interpolation In the next step we will dig deeper into approaches for irregular image interpolation with focus on geometric regularities ACKNOWLEDGEMENT We would like to thank Prof Alain Merigot of Institute de Electronique Fondamentale (lEF), CNRS, France, for providing the high-speed camera sequences used in our experiments, and for his valuable comments on our work [1] Isaac Amidror Scattered data interpolation for electronic imaging systems: a survey Journal of Elecironic Imaging, 11(2): 157-176, 2002 [2] S Baker and T Kanade Super-resolution optical flow Robotics Institute, Carnegie Mellon Univ., Pittsburgh, PA, CMU-Rl-rR'99-36, 1999 [3] JL Barron, DJ Fleet, and SS Beauchemin Performance of optical flow techniques Intemational journal of computer vision, 12(l):43-77, 1994 [4] A Buades, B Coll and J.M Morel A non-local algorithm for image denoising In IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005 CVPR 2005, volume 2, 2005 [5] C Crutchfield Improving Super-Resolution Enhancement of Video by using Optical Flow [6] S Farsiu, M Elad R Milanfar, et al Multiframe dcmosaicing and super-resolution of color images IEEE Transactions on Image Processing, 15C1):141-159, 2006 [7] R Franke Scattered data interpolation: tests of some method Mathematics of Computation, 38C157):181-200 1982 [8] Ha Vu Le and Guna Seetharaman A Super-Resolution Imaging Method Based on Dense Subpixel-Accurate Motion Fields The Journal of VLSI Signal Processing, 42(l):79-89, 2006 [9] A Letienne, F Champagnat, G Le Besnerais, C Kulcsar, and P.V De Lesegno Fast super-resolution on moving objects in video sequences In EUSIPCO European Signal Processing Conference, 2008 [ I | S.H Lim and A El Gamal Optical flow estimation using high frame rate sequences In Image Processing, 2001 Proceedings 2001 International Conference on, volume 2, 2001 [11] C A Micchelli Interpolation of scattered data: distance matrices and conditionally positive definite functions Constructive Approximation, ( I ) : l l - 2 , 1986 [12] S.C Park, M.K Park, and M.G Kang Super-resolution image reconstruction: a technical overview IEEE signal processing magazine, 20(3):21-36, 2003 [13] M Protter, M Elad, H Takeda, and P Milanfar Generalizing the non-local-means to super-resolution reconstruction IEEE Transactions on Image Processing, 18(1):3651 2009 [14] H Takeda Kernel regression for image processing and reconstruction PhD thesis, Citeseer 2006 [15] A Van Eekeren K Schutte, J Dijk DJJ de Lange and LJ van Vliet Super-resolution on moving objects and background In 2006 IEEE Intemational Conference on Image Processing, pages 2709-2712, 2006 [16] X Zhang and X Wu Image interpolation by adaptive 2d autoregressive modeling and soft-decision estimation IEEE Transactions on Image Processing, 17(6):887-896, 2008 [17] W Y Zhao and H.S Sawhney Is super-resolution with optical flow feasible? Lecture Notes in Computer Science, pages 599-613, 2002 I^Al i[()C QUOC GIA HA NOI ruiKJNG DAI HOC C N ( ; NGUK CONC HOA XA \\0\ CllV NCIlf A VJKT NAM , j^p ^ , y ^,^, „ ,j^„|^ ^5^^^ DftciroNC] OK TAI KHOA HOC CONC NCIIE CAP DAI H O C Q U O C CIA HA NOI/CAP I R U O N C DAI HOC CONC NCHfi NAM 2009 (Do Truoiia DHCN quan ly) J I cn dc tai: Tieng l-'iel: Tao dung anh hicn vi voi plian gjai sieu cao phuc vu clio cac ung dung khoa lioc Tieng Anh: appHcations Snpcr-rcsohilion microscopic image construction lor scicnlif ic ThcVi gian thirc hicn: 12 Ihang (lir thang 3/2009 dC'n thang 2/2010) Dc tai Ihuoc ITnh viic uu tien: Xu ly ihoim tin I^inh doc dao (oring quan Xiiy dung de cuong nghien ciru chi tiet 2/2009 3/2009 Dc' cuong chi tiet Chuyen de 2/2009 3/2009 Chuyen di: 2/2000 3/2009 Chuyen de 2/2009 3/2009 3/2009 9/2009 Chuyen de 3/2009 9/2009 Chu}'en de 3/2009 9/2009 Chuyen de 3/2009 9/2009 Xir ly ket quii ngliien ciru thuc nghiem 7/2009 1/2009 Vict ciic bao CcTo chuyen de 1/2009 12/2009 Bo sung so lieu/thu nghiem/irng dung 1/2009 12/2009 Tong kc:l so lieu, ket qua 1/2009 Viet biio cao tong hpp 12/2009 12/2009 12/2009 Moan thien biio cao 12/2009 12/2009 N()p sim phiim 12/2009 12/2009 Nghiem thu de tiii 12/2009 12/2009 D\cu tra khiio siil lam thi nghiem, ihu thap so lieu Ciie biio cao chuyen de Biio eiio long lipp 19 Phan bo kinh phi Ti Noi dim" Kinh phi (VND) Xay dung dc cuong nghien ciru chi tiet 500.000 I T u thiip tiii heu vii vicl long quan vc de Uii 500.000 I1iu thap tiii lieu (mua, thuc') Dich tiii lieu tham khiio (so trang x giii) Viet long quan n J Dic:u tra, khiio sin thi nghiem, thu thap so lieu nghiC-n ciru 500.000 13.500.000 Chi phi lim xe, e(")ng liic phi cho hoat dpng nghien ciru (113) Chi phi th("mg lin lien lac (111) Chi phi thue muon (thue nhim e(")ng, thue chuyen gia hoac ngoiii nuoc) (114) 12.000.000 Chi phi hoat dpng chuyi:n mon: chi to chirc seminar, chi quan ly cua chii nhiem de lai v.v ( 19-99) 1.500.000 Thue, mua sam trang thiet hi, nguyen vat lieu, linh kien 1.500.000 Thuc trang lhi6t bi (119) Mua trang thi^t bi (145) Mua vai lieu, linh kien nho (1 19-01) Viet bao cao khoa hpc, nghiem thu Vict bin bao biio cao dc' tiii loi thao (phi tham du hpi thao, hoi nghi, kinh phi di lai, cong lire phi tham du hoi thao, hoi nghi) (113 hoiic 19-06) 1.500.000 11.500.000 500.000 10.OOO.000 Nghiem thu (1 19-99} 1.000.000 Chi khiic 2.500.000 Mua \'an phiing pliiim ( 19-06) 500.000 in im, cluip tai lieu ( 19-06) 500.000 Quan ly phi (1 19-99) 1.500.000 Tdng kinh phi 30.000.000 Chu thich: Tu} theo diic dicm chuven mon cua lung dc Uii cac muc \a ticu muc Ivowd bang cim CO nhung thav doi cho phu hop viri noi dung ciic hoal dong tighlcn cuu nia so eiic iniie chi (ron" miofic do'n lii dc tham khao lam dir tru kinh pj-ii cho phu hop \a lluian lien DANH MUC CAC TA] LIEU THAM KHAO DT: VII-:1' D|-i CU'ONG (Trinh biiy theo miiu qui dinh ciia ^Tap chi Khoa hpc DHQCiTIN) Tiii lieu tieng Vicl Tiii lieu tieng nuoc ngoiii S.C Park, M.K Park, and M.G Kang '^ Super-resolution image reconstruction: a technical overview", IEEE Signal Processing Magazine, Vol.20., No.3 May 2003 p a g e s D S Baker and T Kanade ^M.amils on Super-Resolution and How to Break Them" IEEE I'ransaclions on Fallern Analysis and Machine Inledigoice Vol.24, No.9, September 2002, pages 167-1 183 M Elad and A Feuer " Restoration o k a single superresolution image Trom several blurred, Transactions noisy, and undersampled on Image Processing measured images" //:/:7i Vol.6 No I 2, December 1997, pages 1646- 1658 Y Yitzhaky and N S Kopeika, ^•Identilieation okf^lur Paramelers from Motion Blurred Images'" Graphical Models and Image Proeessing Vol.59, No.5, September 1997 pages 10-320 Neiiv 20 Ihiiim nam 2008 Thii l i m m g don \ i Chu nhiem de liii (Kiioa hoac Truny lam line liiiioc) \iic nhim \ \% Vu Tw /p 1^1 TRl»J(^-DAl HOt CONG NGHE' / N u 11 t h a n , ^ nan i i ^ SAO DUNG ^^^^ CHINH Naav/.Mhana .^t '"'^''—'Wwmmi O'l Phe duN'cl cua T r u o i m J ) l ICN ' T / L HIEU TRUONG TRUONG PHONG ^ f """ Afl!ONGI(klll]C-HANH':H!flH yyiMf^^sL '^f DAI HOC QUOC GIA HA N O I TRUONG DAI HOC CONG NGHE CQNG HOA XA HOI CHU NGHIA VIET NAM Doc lap - Tu - Hanh phuc S6: ^ f / H D - N C K H Ha Noi, ngdy3.0 thdng nam 2009 HOP DONG THI/C HIEN DE TAI NGHIEN ClTU KHOA HOC CAP DAI HOC QUOC GIA HA N Q I NAM 2009 - Can cu ve Qui dinh ve To chuc vo Hoal dong cua Dgi hoc QuSc gia Ha Noi ban hanh theo Quyel dinh so 600/TCCB OJ ihang ID nam 2001 cua Dgi hoc Qudc gia Ho Noi qui dinh quyen hgn cua hien trucmg cdc trudng dgi hoc vien: - Cdn cit- Thong boo sd J97J/TB-KHCN 03 thdng nam 2009 cua Gidm dSc Dgi hoc Qudc gia Ho Npi ve viec Giao nhiem vit va chi tieu ki hooch khoa hoc & cong nghe nam 2009; - Can cu de cuong nghien ciru ciio de tai da duac phe duyet, Chiing loi gom: Ben giao nhiem vu (goi la ben A): Truong Dai hoc Cong nghe - DHQG Ha Noi Dai dien la: PGS TS NguySn Ngoc Binh Chirc vu: Hieu truang Ben nhan nhiem vu (goi la ben B) Ong: TS Le Vu Ha Don vi cong tac: Klioa Dien tii' Vien thong - Truang Dai hpc Cong nghe Ky hop dong thuc hien de tai nghien ciru khoa hpc cap E^ai hpc Quoc gia Ha Npi'' Ten de tai: "Tao dung anh hien vi vol phan giai sieu cao phuc vu cho cac ling dung khoa hoc" Ma so: QC.09.07, Vai nhung dieu khoan thoa thuan nhu sau: Dieu ] : Ben B chiu trach nhiem to chirc trien khai thuc hien cac npi dung nghien ciiu ciia de lai theo diing tien dp thuc hien da dang ky de cuang nghien cii'u da dupe phe duyet Dieu 2: Ben B bao cao ket qua thuc hien de tai va giao nop cae san pham ciia de tai cho ben A theo dimg cac qui dinh hien hanh cua Dai hpc Quoc gia Ha Npi va ciia Truang Dai hpc Cong nghe truac 20/06/2010, bao gom: 01 bai bao dang tai ky yeu hoi nghj quoc te Tong quan ve de tai kem theo file dien tii (Mot ban bang tieng Viet, mpt ban bang tieng Anh - Highlight; moi ban diii khoang 400 tir tren mpt trang giay kho A4, font Time New Roman, co chCr 13pl- each dong dan: Npi dung: Tom tat muc tieu, phuang phap va npi dung nghien ciru, ket qua dat dupe, danh gia y nghTa va tac dpng khoa hpc cong nghe ciia cac ket qua dat dupe cung nhu ciia viec thue hien de tai) fBdng chii: Ba muoi trieu dong chan) Chi phi cu the nhu du toan cua ban du trii kinh phi ^^^^Jl Ben B co trach nhiem su dung kinh phi duac cdp theo dimg muc dich dung che dp tai chinh hien hanj-i, quy^t toan vai phong Tai vu - K^ toan \'a thirc hien viec nghiem thu de tai theo dimg qui dinli ciia Dai hpc Qu6c gia Ha Npi Djeu^S: Ben A giu quy^n so him tri tue d6i voi cac k^t qua khoa hpc ciia de tai fat ca cae cong bo hen quan d^n noi dung khoa hpc cua d^ tai phai ghi ro ngu6n tai tro kinh phi nghien ciru theo ma so ciia d^ tai nhu sau: - Doi vai bill bao, bao cao kJioa hpc: "Cong trinh dupe tai tra mpt phan tir de tiii mang ma so: QC.09.07 Dai hpc Qa6c gia Hii Npi" - Doi vai luan van (khoa luan ): "Luan van (khoa luan } dupe thuc hien khuon kho de tai mang ma so: QC.09.07 Dai hpc Quoc gia Hii Npi Doi voi bai bao, bao cao dang a tap chi, ky y^u hpi nghi quoc te (tieng Anh): "This vNork is (panly) supported by the research project No QC.09.07 granted h> Vietnam National University Hanoi"' Dieu 6: Hai ben cam ket thuc hien diing cac dieu khoan da ghi hop dong Trong qua triuli thue hien hop dong hai ben co trach nhiem thong bao kip thai cho nliau nhirng \-an de vuang mac va cimg ban bac, tich cue tim bien phap giai quyet Hpp dong tu dpng duoc ly sau c6 bien ban hpp hpi dong khoa hpc danh gia nghiem thu de tiii \ai ket qua dap img cac qui dinh hien hanh Dieu 7: Hop dong lam 05 ban moi ben giir 01 ban 02 ban gui cho phong TV-tCf 01 ban luu tai phong TC-HC IEN BEN A uyen Ngoc Binh DAI DIEN BEN B TS Le Vii Ha MAU 26/KHCN (Kem theo Quyit dinh s6 I895/QD-DHQGHN ciia Gidm dSc DHQGHN) ngdy 24/6/2010 BAG CAO TOIM TAT BANG TIENG ANH KET QUA THlTC HIEN DE TAI KHCN CUA DHQGHN SUMMARY Project Title: Super-resolution micrscopic image construction for scientific applications Code number: QC.09.07 Coordinator: Le Vu Ha Implementing Institution: University of Engineering and Technology, VNU Hanoi Cooperating Institution(s): Duration: from June 2009 To June 2010 Objectives: to evaluate the performance of methods for constructing super-resolution images from optical microscopic image sequences Main contents: - Motion estimation methods for super-resolution image construction from image sequences - Image interpolation methods for super-resolution image construction from image sequences - Experimental results Results obtained: - Publication: Hong-Thinh Nguyen, Viet-Dung Nguyen, Ha Vu Le, ''Super-Resolution Image Construction from High-Speed Camera Sequences", Proceedings of the Intemational Conference on Advanced Technologies for Communications, Ho Chi Minh City, October 2010 - 01 undergraduate thesis, 01 master's thesis - Software tools for constructing super-resolution images from image sequences Signature MAU 27/KHCN (Kem theo Quyet dinh s6 I895/QD-DHQGHN 24/6/2010 cua Gidm doc DHQGHN) PHIEU DANG KY KET QUA NGHIEN CU'U CUA DE TAI, DlT AN KHCN Ten de tai, dir an: Tao dung anh hien vi vai phan giai sieu cao phuc vu cho cac img dung khoa hoc Ma so: QC.09.07 Co quan chu tri de tai: Trucmg Dai hoc Cong nghe Dia chi: 144 Xuan Thiiy, C£iu Gidy, Ha Noi Tel: Ca quan quan ly de tai: Dai hoc Quoc gia Ha Noi Dja chi: 144 Xuan Thuy, Cau GiAy, Ha Noi Tel: Tong kinh phi thirc hien: 30 trieu dong Trong do: Tir ngan sach nha nuac: 30 trieu dong Thai gian nghien cuu: 12 thang - Thai gian bdt dau: 6/2009 - Thai gian k^t thuc: 6/2010 Ten cac can bo phoi hap nghien ciru: - LeVuHa Nguyen Thang Long So dang ky de tai: Ngay: Tom tat ket qua nghien cim: - Thir nghiem va danh gia mot so phuang phap tao anh sieu phan giai nham ap dung cho anh hien vi quang hoc - Xay dimg cong cu phan mem tao anh sieu phan giai Cdc san phdm khoa hoc va dao tao: 01 khoa luan dai hoc 01 luan van thac sT 01 cong trinh NCKH cua SV dat giai nhit cap truang va giai khuyen khich cdp Bo GD-DT Cdc bdi bdo da duac cong bd tren cdc tap chi khoa hoc: Hong-Thinh Nguyen, Viet-Dung Nguyen, Ha Vu Le, "Super-Resolution Image Construction from High-Speed Camera Sequences", Proceedings of the Intemational Conference on Advanced Technologies for Communications, Ho Chi Minh City, October 2010 Kit qud tham gia dao tao Sau Dgi hoc: 01 luan van thac sT Kien nghi ve qui mo va doi Urgng ap dung nghien ciru: 4W^^'fe/