Workshop proceeding - final.pdf - Faculty of Information and ...

More documents

Recommendations

Info

Where u ′ , v′ ) is the point position in the full resolution image Im Oe , ρ u ′ , v′ ) is the correlation ( k k ( k k coefficient of accurate image registration, ρl ( uk , vk ) is the correlation coefficient of rough image registration. Through the step, we can get the registration position u ′ , ′ ) . ( * v * k k The final step is the estimation of rotation angle. At present the permitted rotation angle θ range is from -10°to 10°. We divide the range into 200 angles. To each angle, we calculate its correlation coefficient, then choose the angle θ * with the maximum correlation value. 4. Experiment results The above algorithm was tested using many image pairs. All of the results are successful with tolerable errors(Fig 3). In order to test the running speed, contrasting experiment was done between the changeable resolution algorithm and standard NCC in the same running environment and conditions, the results can be shown in table 1. The experiments show that the changeable resolution algorithm has the same good registration results with a fast speed, only spending less than one fortieth running time compared with the standard algorithm. Because we don’t know the true registration positions, we compare them with the results of manual registration. Fig 3 Raw image pairs and their registration results Table 1 experiment results of changeable resolution algorithm based on NCC Number of Success Running time per pair Registration error Image pairs number (average) (max, min, average) Standard NCC 20 20 610s (6,0,2.6) Changeable resolution algorithm 20 20 15s (6,0,2.6) All of the tests are done in the condition of f = 2 and using Sobel operator to get the image edges. The contrasting experiments with the NMI (normalized mutual information) and CR(correlation ratio) algorithms have also been done( Fig 4, Table 2). From these figures and tables, we can see the changeable resolution algorithm based on normalized cross correlation is the best choice. 30
Table 2 experiment results of several different algorithms Number of Image Success* Registration error pairs number (max, min, average) Changeable resolution 10 10 (6,0,2.6) algorithm NMI algorithm 10 8 (9,0,2.7) CR algorithm 10 0 Null Remark: *here a successful registration is whose position error is less than 10 pixels. (a) Image pair 1(Optical(left), IR(right)) (299,120,-0.8) (b) registration result 1 of Changeable Res Algo (c) registration result 1 of NMI alg (304,125,-0.5) (d) registration result 1 of CR Alg (318,257,0.5) (e) Image pair 2(Optical(left), IR(right)) (f) registration result 2 of Changeable Res Algo (364,153,1.5) (g) registration result 2 of NMI alg (363,157,1.0) (h) registration result 2 of CR Alg (283,93,3.9) Fig 4 Raw image pairs and their registration results using different algorithms 5. Discussion From the experiment results we can find that NMI and CR algorithms are unsuitable to the different source image registration because these two methods are dependent on the intensity distributions of the images, while the changeable resolution algorithm is related to the edge of the images, that is to say, the latter algorithm is dependent on the image features and independent with the image intensities. Though they have good results in many applications, the image registration methods 31
Page 2 and 3: The 3 rd VACPS Research Workshop Pr
Page 4 and 5: Organizing Committee Victorian Asso
Page 6 and 7: Preface On the occasion of the publ
Page 8 and 9: Paper as a low-cost base material f
Page 10 and 11: Acknowledgements We gratefully ackn
Page 13 and 14: Career planning and Job Interview i
Page 15 and 16: The role of mitochondria in atrial
Page 18 and 19: Track-Before-Detect Procedures for
Page 20 and 21: 3. Dynamic model for approaching ta
Page 22 and 23: ⎧π0 if j = i and Cx ij k > δij
Page 24 and 25: k k k k ( ) p( p Ζ | X , B = ∏ z
Page 26 and 27: As illustrated in Fig.4, the struct
Page 28 and 29: ( xˆ , m ˆ ) = argmax Γ ( x , m
Page 30 and 31: (a) (b) Fig. 7. mesh plot of the re
Page 32 and 33: Definition 2 The probability of tar
Page 34 and 35: Acknowledgments Xiaobo Deng would l
Page 36 and 37: speed, several methods can be utili
Page 40 and 41: ase on SIFT are also unsuitable to
Page 42 and 43: Demonstration and Performance Analy
Page 45 and 46: Effect of mini-tyrosl-tRNA syntheta
Page 47 and 48: 2.1 Materials Sprague-Dauley(SD) ma
Page 49 and 50: Fig. 2. The representative photomic
Page 51 and 52: mount/0.1mm 2 , as for mini-TrpRS g
Page 53 and 54: 3.4.The mRNA expression of mini-Tyr
Page 55 and 56: 9.Du XJ, Gao X, Ramsey D. Surgical
Page 57 and 58: Distribution of three forms alpha-m
Page 59 and 60: Materials and Methods Ethics All an
Page 61 and 62: Fig.1 (A) Mass Spectrometry results
Page 63 and 64: indicated that there is more des-α
Page 65 and 66: Fig.6 (A) Mass Spectrometry results
Page 67 and 68: transcription-polymerase chain reac
Page 69 and 70: Investigation of the mechanism of t
Page 71 and 72: Dual mode roll-up effect in a multi
Page 73: Paper as a low-cost base material f
Page 76 and 77: cloud computing offers a cost-effec
Page 78 and 79: workflows. By comparing the generat
Page 80 and 81: [10] E. Deelman, G. Singh, M. Livny
Page 82 and 83: Web Page Prediction Based on Condit
Page 84 and 85: A Multi-modal Gesture Recognition S
Page 86: Searching for Fair Joint Gains in A
Page 89 and 90:
environments. However, challenges a
Page 91 and 92:
conclude that the hydrophobic PHB c
Page 93 and 94:
Figure 4. (a) Energy transfer model
Page 95 and 96:
20 Intensity (/10 6 ) / a.u. 15 10
Page 97 and 98:
(22) (a) Cho, J.; Quinn, J. F.; Car
Page 99 and 100:
Grain refinement of pure magnesium
Page 101 and 102:
Fig. 3 The optical and SEM microstr
Page 103 and 104:
Conclusions (1) The large grains in
Page 105 and 106:
[22-25]. Sorting of normal and Babe
Page 107 and 108:
Figure 1. Г×Re[fCM] vs. frequency
Page 109 and 110:
Figure 3. The DEP spectrum of cylin
Page 111 and 112:
A B C Figure 5. The gradient of the
Page 113 and 114:
Figure 7. SEM image of damaged elec
Page 115 and 116:
Figure 9. Particle separation progr
Page 117 and 118:
The separation efficiency can be im
Page 119 and 120:
[37] A. Motayed, M. Q. He, A. V. Da
Page 121 and 122:
Fig. 1. The transverse and longitud
Page 123 and 124:
References: [1] Xiong Liu, Mark C.
Page 125 and 126:
new tissue[10]. Most of these human
Page 127 and 128:
2.8 Viability and morphology study
Page 129 and 130:
Fig 3 and 4 SEM images of electrosp
Page 131 and 132:
studies have indicated that surface
Page 133 and 134:
Fig 10 Tensile Stress curve of elec
Page 135 and 136:
well as on a usual nanofibrous scaf
Page 137 and 138:
Uniform Coating of WO x on TiO 2 Na
Page 140 and 141:
Union as a Social Regulator of Mark
Page 142 and 143:
associated with a reinvention of in
Page 144 and 145:
enterprises…There obviously is no
Page 146 and 147:
Union assistants participate in org
Page 148 and 149:
[5] Deyo, Frederic C and Ağartan,
Page 150 and 151:
Organizational Capabilities • Lea
Page 152 and 153:
Table 1. Correlations, Reliabilitie
Page 154 and 155:
[18]. Jap, S. D. (1999). Pie-expans
Page 156 and 157:
Early Childhood Education Matters:
Page 158 and 159:
decisions on selection of child car
Page 160 and 161:
Noble, K. (2007). Parent choice of
Page 162:
An Exploration of Country of Origin
Page 165 and 166:
Supercontinuum generation for fiber
Page 167 and 168:
Gas chromatographic retention indic
Page 169 and 170:
Outward Foreign Direct Investment f
Page 171 and 172:
Fiber Nonlinearity Precompensation
Page 173 and 174:
Improved Nonlinearity precompensati
show all

Workshop proceeding - final.pdf - Faculty of Information and ...

Create successful ePaper yourself

Delete template?

Save as template?