175:
resolution by inferring some high-frequency face details from the low-frequency facial information by taking advantage of the correlation between the two parts. Because of the structural similarity among face images, in multiresolution analysis, there exists strong correlation between the high-frequency band and low-frequency band. For high-resolution face images, PCA can compact this correlated information onto a small number of principal components. Then, in the eigentransformation process, these principal components can be inferred from the principal components of the low-resolution face by mapping between the high- and low-resolution training pairs.
171:. This method sees the solution as a transformation between different styles of image and uses a principal component analysis (PCA) applied to the low-resolution face image. By selecting the number of "eigenfaces", we can extract amount of facial image information of low resolution and remove the noise.
112:
The simplest way to increase image resolution is a direct interpolation increasing the pixel intensities of input images with such algorithms as nearest-neighbour, bilinear and variants of cubic spline interpolation. Another approach to interpolation is to learn how to interpolate from a set of high
203:
The method is presented in three-step framework. Firstly, a low-resolution input image is up-sampled by an interpolation. The interpolated image can be represented as a superposition of the global high-resolution image and an “unsharp mask”. In the second step, the interpolated image is decomposed
43:
is a class of techniques that enhance the resolution of an image using a set of low resolution images. The main difference between both techniques is that face hallucination is the super-resolution for face images and always employs typical face priors with strong cohesion to face domain concept.
191:
In first step, learn the relationship between the high resolution image and their smoothed and down-sampled. In second step, model the residue between an original high resolution and the reconstructed high-resolution image after applying learned lineal model by a non-parametric Markov network to
174:
In the eigentransformation algorithm, the hallucinated face image is synthesized by the linear combination of high-resolution training images and the combination coefficients come from the low-resolution face images using the principal component analysis method. The algorithm improves the image
73:(MAP). The second step produces residual image to compensate the result of the first step. Furthermore, all the algorithms are based on a set of high- and low-resolution training image pairs, which incorporates image super-resolution techniques into facial image synthesis.
56:
Moreover, the challenge in face hallucination is the difficulty of aligning faces. Many methods are required to bring the alignment between the test sample taken and the training samples. Even a slight amount of wrong alignment can degrade the method and the result.
204:
into a global high-resolution image by using MCA to obtain the global approximation of the HR image from interpolated image. Finally, facial detail information is compensated onto the estimated HT image by using the neighbour reconstruction of position-patches.
52:
An image is considered high resolution when it measures 128Ă—96 pixels. Therefore, the goal of face hallucination is to make the input image reach that number of pixels. The most common values of the input image is usually 32Ă—24 pixels or 16Ă—12 pixels.
151:
This method was proposed by J. Yang and H. Tang and it is based in hallucinating of High-Resolution face image by taking Low-Resolution input value. The method exploits the facial features by using a Non-negative Matrix factorization
65:
In the last two decades, many specific face hallucination algorithms have been reported to perform this technique. Although the existing face hallucination methods have achieved great success, there is still much room for improvement.
26:
technique which applies specifically to faces. It comprises techniques which take noisy or low-resolution facial images, and convert them into high-resolution images using knowledge about typical facial features. It can be applied in
127:
The algorithm is based on
Bayesian MAP formulation and use gradient descent to optimize the objective function and it generates the high frequency details from a parent structure with the assistance of training samples.
766:
183:
This method was developed by C. Liu and Shum and it integrates a global parametric and a local parametric model. The global model is a lineal parametric inference and the local model is a patch-based non-parametric
740:
95:
The output image must have very specific features of the face image having resemblance with photorealistic local features. Without this constraint, the resulting image could be too smooth.
238:
All methods presented above have very satisfactory results and meet expectations, so it is difficult to determine which method is most effective and which gives a better result.
508:
31:
for identifying faces faster and more effectively. Due to the potential applications in facial recognition systems, face hallucination has become an active area of research.
89:
The resulting image always contains all common features of a human face. The facial features must be coherent always. Without this constraint, the output could be too noisy.
744:
636:
669:
69:
The common algorithms usually perform two steps: the first step generates global face image which keeps the characteristics of the face using probabilistic method
776:
372:
143:(PCA) basis and reconstructs the area separately. However, the reconstructed face images in this method have visible artifacts between different regions.
200:
This algorithm formulates the face hallucination as an image decomposition problem and propose a
Morphological Component Analysis (MCA) based method.
791:
116:
However, the results are very poor since no new information is added in the process. That is why new methods have been proposed in recent years.
564:
694:
341:
713:
677:
Zhen Jia; Hongcheng Wang; Ziyou Xiong; Finn, Alan (2011). "Fast face hallucination with sparse representation for video surveillance".
515:
153:
610:
771:
70:
139:
It divided the face image into four key regions: the eyes, nose, mouth and cheek areas. For each area, it learns a separate
140:
811:
806:
156:) approach to learn localized part-based subspace. That subspace is effective for super-resolving the incoming face.
113:
resolution training samples, together with the corresponding low resolution versions of them. (pg 4 baker and kanade)
28:
159:
For further enhance the detailed facial structure by using a local patch method based on sparse representation.
320:
Proceedings of the 2001 IEEE Computer
Society Conference on Computer Vision and Pattern Recognition. CVPR 2001
168:
767:
Hallucinating Faces: Global Linear Modal Based Super-Resolution and
Position Based Residue Compensation
405:
282:
387:
591:
611:"A two-step approach to hallucinating faces: Global parametric model and local non-parametric model"
482:
104:
Face hallucination enhances facial features with improved image resolution using different methods.
571:
310:
720:
700:
366:
347:
741:"Hallucinating faces: LPH super-resolution and neighbor reconstruction for residue compensation"
690:
663:
630:
459:
337:
124:
This method was proposed by Baker and Kanade, the pioneering of face hallucination technique.
682:
329:
83:
The output image should be nearly to the original image when it is smoothed or down-sampled.
786:
433:
418:
295:
40:
23:
645:
614:
245:
The method of Baker and Kanade can distort the characteristic features of a face image.
185:
325:
318:
800:
213:
Face hallucination by tensor patch super-resolution and coupled residue compensation.
646:"Hallucinating Faces: TensorPatch Super-Resolution and Coupled Residue Compensation"
351:
704:
267:
781:
545:
686:
248:
The result of the method developed by Wang and Tang can create ringing effect.
136:
Capel and
Zisserman was the first to propose the local face image SR method.
333:
229:
LPH super-resolution and neighbor reconstruction for residue compensation.
76:
Any face hallucination algorithm must be based in three constraints:
311:"Super-resolution from multiple views using learnt image models"
772:
Neighbor
Combination And Transformation For Hallucinating Faces
509:"Face Hallucination based on Morphological Component Analysis"
132:
Super-resolution from multiple views using learnt image models
167:
This method was proposed by Wang and Tang and it uses an
35:
Differences between face hallucination and super-resolution
609:Ce Liu, Heung-Yeung Shum, Chang-ShuiZhang (2013).
386:Yang, Jianchao; Tang, Hao; Ma, Yi; Huang, Thomas.
317:
679:The First Asian Conference on Pattern Recognition
192:capture the high-frequency content of faces.
8:
635:: CS1 maint: multiple names: authors list (
590:Chih-Yuan Yang; Sifei Liu; Ming-Hsuan Yang.
739:Yueting Zhuang; Jian Zhang; Fei Wu (2007).
668:: CS1 maint: numeric names: authors list (
434:"Hallucinating Face by Eigentransformation"
714:"Position-Based Face Hallucination Method"
546:"Differents Methods of Face Hallucination"
371:: CS1 maint: location missing publisher (
483:"Face Hallucination: Theory and Practice"
460:"Face Hallucination: Theory and Practice"
163:Face Hallucination by Eigentransformation
120:Face hallucination based on Bayes theorem
506:Yan Liang, Xiaohua Xie, Jian-Huang Lai
324:. Vol. 2. Kauai, Hawaii. pp.
258:
661:
628:
414:
403:
388:"Face Hallucination Via Sparse Coding"
364:
291:
280:
644:Wei Liu1, Dahua Lin and Xiaoou Tang.
223:Hallucinating face by position patch.
16:Aspect of facial recognition software
7:
147:Face Hallucination via Sparse Coding
712:Xiang Ma; Junping Zhang; Chun Qi.
480:C. Liu, H.Y. Shum and W.T Freeman
457:C. Liu, H.Y. Shum and W.T Freeman
14:
309:Capel, D.; Zisserman, A. (2001).
592:"Estructured Face Hallucination"
241:However, it can be stated that:
196:Face hallucination based on MCA
565:"Face Hallucination: A Review"
431:Xiaogang Wang and Xiaoou Tang
1:
266:Baker, Simon; Kanade, Takeo.
141:Principal Component Analysis
563:Kaur, Jaskiran (May 2014).
544:Kaur, Ravneet (June 2014).
828:
782:Eigenfaces for recognition
514:. Oct 2012. Archived from
29:facial recognition systems
687:10.1109/ACPR.2011.6166702
334:10.1109/CVPR.2001.991022
220:for video surveillance.
413:Cite journal requires
290:Cite journal requires
268:"Hallucinating Faces"
218:sparse representation
216:Superresolution with
681:. pp. 179–183.
71:maximum a posteriori
747:on 30 November 2014
169:eigentransformation
812:Facial recognition
807:Identity documents
726:on 5 December 2014
617:on 5 December 2014
577:on 5 December 2014
521:on 5 December 2014
20:Face hallucination
696:978-1-4577-0121-4
343:978-0-7695-1272-3
179:Two-step approach
86:Global constraint
819:
756:
754:
752:
743:. Archived from
735:
733:
731:
725:
719:. Archived from
718:
708:
673:
667:
659:
657:
655:
650:
640:
634:
626:
624:
622:
613:. Archived from
605:
603:
601:
596:
586:
584:
582:
576:
570:. Archived from
569:
559:
557:
555:
550:
531:
530:
528:
526:
520:
513:
504:
498:
497:
495:
493:
487:
478:
472:
471:
469:
467:
455:
449:
448:
446:
444:
438:
429:
423:
422:
416:
411:
409:
401:
399:
397:
392:
383:
377:
376:
370:
362:
360:
358:
323:
315:
306:
300:
299:
293:
288:
286:
278:
276:
274:
263:
92:Local constraint
827:
826:
822:
821:
820:
818:
817:
816:
797:
796:
763:
750:
748:
738:
729:
727:
723:
716:
711:
697:
676:
660:
653:
651:
648:
643:
627:
620:
618:
608:
599:
597:
594:
589:
580:
578:
574:
567:
562:
553:
551:
548:
543:
540:
535:
534:
524:
522:
518:
511:
507:
505:
501:
491:
489:
485:
481:
479:
475:
465:
463:
458:
456:
452:
442:
440:
436:
432:
430:
426:
412:
402:
395:
393:
390:
385:
384:
380:
363:
356:
354:
344:
313:
308:
307:
303:
289:
279:
272:
270:
265:
264:
260:
255:
236:
226:Position-based.
210:
208:Another methods
198:
181:
165:
149:
134:
122:
110:
102:
80:Data constraint
63:
50:
41:superresolution
37:
24:superresolution
17:
12:
11:
5:
825:
823:
815:
814:
809:
799:
798:
795:
794:
789:
784:
779:
774:
769:
762:
761:External links
759:
758:
757:
736:
709:
695:
674:
641:
606:
587:
560:
539:
536:
533:
532:
499:
488:. October 2007
473:
462:. October 2007
450:
424:
415:|journal=
378:
342:
301:
292:|journal=
257:
256:
254:
251:
250:
249:
246:
235:
232:
231:
230:
227:
224:
221:
214:
209:
206:
197:
194:
186:Markov network
180:
177:
164:
161:
148:
145:
133:
130:
121:
118:
109:
106:
101:
98:
97:
96:
93:
90:
87:
84:
81:
62:
59:
49:
46:
36:
33:
22:refers to any
15:
13:
10:
9:
6:
4:
3:
2:
824:
813:
810:
808:
805:
804:
802:
793:
790:
788:
785:
783:
780:
778:
775:
773:
770:
768:
765:
764:
760:
746:
742:
737:
722:
715:
710:
706:
702:
698:
692:
688:
684:
680:
675:
671:
665:
647:
642:
638:
632:
616:
612:
607:
593:
588:
573:
566:
561:
547:
542:
541:
537:
517:
510:
503:
500:
484:
477:
474:
461:
454:
451:
435:
428:
425:
420:
407:
389:
382:
379:
374:
368:
353:
349:
345:
339:
335:
331:
327:
322:
321:
312:
305:
302:
297:
284:
269:
262:
259:
252:
247:
244:
243:
242:
239:
233:
228:
225:
222:
219:
215:
212:
211:
207:
205:
201:
195:
193:
189:
187:
178:
176:
172:
170:
162:
160:
157:
155:
146:
144:
142:
137:
131:
129:
125:
119:
117:
114:
108:Interpolation
107:
105:
99:
94:
91:
88:
85:
82:
79:
78:
77:
74:
72:
67:
61:The algorithm
60:
58:
54:
47:
45:
42:
34:
32:
30:
25:
21:
787:Unsharp Mask
749:. Retrieved
745:the original
728:. Retrieved
721:the original
678:
652:. Retrieved
619:. Retrieved
615:the original
598:. Retrieved
579:. Retrieved
572:the original
552:. Retrieved
538:Bibliography
523:. Retrieved
516:the original
502:
490:. Retrieved
476:
464:. Retrieved
453:
441:. Retrieved
427:
406:cite journal
394:. Retrieved
381:
355:. Retrieved
319:
304:
283:cite journal
271:. Retrieved
261:
240:
237:
217:
202:
199:
190:
182:
173:
166:
158:
150:
138:
135:
126:
123:
115:
111:
103:
75:
68:
64:
55:
51:
38:
19:
18:
751:26 November
730:26 November
654:26 November
621:17 November
600:20 November
581:16 November
554:16 November
525:21 November
492:20 November
466:20 November
443:17 November
273:18 November
801:Categories
792:Markov net
777:Eigenfaces
253:References
367:cite book
664:cite web
631:cite web
352:14090080
48:Measures
705:9913575
396:4 March
357:4 March
326:627–634
234:Results
100:Methods
703:
693:
439:. 2005
350:
340:
39:Image
724:(PDF)
717:(PDF)
701:S2CID
649:(PDF)
595:(PDF)
575:(PDF)
568:(PDF)
549:(PDF)
519:(PDF)
512:(PDF)
486:(PDF)
437:(PDF)
391:(PDF)
348:S2CID
314:(PDF)
753:2014
732:2014
691:ISBN
670:link
656:2014
637:link
623:2014
602:2014
583:2014
556:2014
527:2014
494:2014
468:2014
445:2014
419:help
398:2015
373:link
359:2015
338:ISBN
296:help
275:2014
683:doi
330:doi
154:NMF
803::
699:.
689:.
666:}}
662:{{
633:}}
629:{{
410::
408:}}
404:{{
369:}}
365:{{
346:.
336:.
328:.
316:.
287::
285:}}
281:{{
188:.
755:.
734:.
707:.
685::
672:)
658:.
639:)
625:.
604:.
585:.
558:.
529:.
496:.
470:.
447:.
421:)
417:(
400:.
375:)
361:.
332::
298:)
294:(
277:.
152:(
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.