You are on page 1of 4

Journal of Computer Applications (JCA) ISSN: 0974-1925, Volume VI, Issue 2, 2013

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques

B.Ramesh a, *, R.Kannan b,1

Abstract - Construction of a 3-D face model from a 2-D face image is fundamentally important for face recognition and animation because the 3-D face model is invariant to changes of viewpoint, illumination, background clutter, and occlusions. Given a coupled training set that contains pairs of 2-D faces and the corresponding 3-D faces, we train a novel Neural network (Feed Forward Back Propagation) to recover the 3-D face model from a single 2-D face image. The particular face can be reconstructed by its nearest neighbors, we can assume that the linear combination coefficients for a particular 2-D face image reconstruction are identical to those for the corresponding 3-D face model reconstruction. Therefore, we can reconstruct a 3-D face model by using 2-D face image based on the neural networks.

Index Terms Reconstruction, Representation, Back Propagation, Multilayer Feed-Forward Neural Networks, Affined Transformation, Third Order Polynomial, Object Space.

I.

INTRODUCTION

Reconstructing a 3D model from a set of images is a popular field of work in computer vision. Scene reconstruction involves building 3-D models of scenes, given several 2-D photographs of the scenes. It has its applications in the field of virtual reality, tele-presence, and visual navigation. 3-D scene reconstruction has been carried out for many years from now and many algorithms have been proposed so far. The improved GVC algorithm that has been proposed takes into the account histogram based consistency checks and edge detection. The Generalized Voxel Coloring of Culbertson and Malzbender, (GVC) generalises Space Carving so that all Views can be considered simultaneously, rather than with separate passes. It removes the camera positioning restrictions that existed in the original voxel coloring algorithm. Any reconstruction algorithm is based on certain rules that specify whether the surfaces or regions shown in the images should be represented in 3-D reconstruction or not. One of the rules is based on color consistency. This paper talks about histogram consistency check that is well suited for characterising multi modal distributions of color sets. This measure has its own advantages over other consistency checks such as the ones based on standard deviation, means and variances.

B.Ramesh, Department of Electronics and Communication Engineering, Er.Perumal Manimekalai College of Engineering, Hosur, India, +91-8438332483., (E-mail:bmramesh4u@gamil.com).

This paper highlights those advantages and discusses this photo consistency check in detail. It proposes a two stage consistency check where at first stage various sets of pixels belonging to the different footprints of a particular voxel are compared for similarity and consistency. In case the consistency holds, voxel is kept in the 3-D model, else it is declared inconsistent. At the second stage, all the voxels that were declared inconsistent at stage one, are re-evaluated for edges and other features to test whether the regions that voxel projects onto contain edges or not. In case it does, voxel is declared consistent, else it is carved out. Finally, all the voxels that are consistent, are kept in the 3-D reconstructed object, others form a part of the background. This paper discusses the improved GVC 2- stage consistency check algorithm in detail and the role of histogram measure in detail.

II. PREVIOUS RELATED WORK

Various consistency measures have been using for image based rendering algorithms in the past. When reconstructing a scene using a reconstruction algorithm, there are two key factors that affect the quality of the reconstructed model. The first is the visibility that is computed for the voxels. The second factor is the test that is used to judge the photo consistency of voxels. The consistency test just can be thought of as being too strict for declaring voxels that belong in the model to be inconsistent. Tests can also be too lenient, declaring voxels to be consistent when they do not belong in the model; this can lead to voxels that appear to float over a reconstructed scene. A single consistency test can simultaneously be both too strict and too lenient, creating holes in one part of a scene and floating voxels elsewhere. In most space carving implementations there has been an implicit assumption that the pixel resolution is greater than the voxel resolutionthat is, a voxel projects to a number of pixels in at least some of the images. We believe this is reasonable and expect the trend to continue because runtime grows faster with increasing voxel resolution than it does with increasing pixel resolution. We make the assumption in this section that the scenes being reconstructed are approximately Lambertian, and we use the RGB color space, except where noted. Most of the color consistency checks require a threshold value. One of the earlier methods used was the method of standard deviation. Using standard deviation as a photo consistency measure is proposed by Seitz and Dyer.

III.

PROPOSED SYSTEM

R.Kannan, Department of Electronics and Communication Engineering, Er.Perumal Manimekalai College of Engineering, Hosur, India, +91-9655885949., (E-mail:kannan.vlsi.06@gamil.com).

We are concerned about only the object, so we don’t want region outside the object. So, we are going to find the edges of the object in the given 2D Image. A feature is defined as an interesting part of an image. It is an most important step

36

Journal of Computer Applications (JCA) ISSN: 0974-1925, Volume VI, Issue 2, 2013 Three-Dimensional Face Reconstruction from

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques

in processing an image. There are many types of features available in an image like edges, corners and ridges. We are going to use corner because edge detection intern uses corner detection as the primary model. Harris corner method is used to detect the corners in the 2D-Image.

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques in processing an image. There

Figure1.Proposed workflow diagram

A. Harris Corner Method

Step1:

Compute x and y derivatives of the 2D-Image. Ix=Gx*I;Iy=Gy*I;

Step2:

Compute product of derivatives at every pixel. Ix 2 =Ix*Ix; Iy 2 =Iy*Iy; Ixy=Ix*Iy;

Step3:

Compute the sums of the products of the derivatives at each pixel;

Sx2=G* Ix 2 Sy2=G*Iy 2 Sxy=G*Ixy;

Step4:

Define at each pixel(x,y) the matrix

H(x,y)= Sx2(x,y)

Sxy(x,y

Sxy(x,y)

Sy2(x,y)

Step5:

Compute the response of the detector at R=Det(H)-K(H) 2

Step6:

Threshold on value of R .If it exceeds the predefined value mark that pixel.

B.Neural Networks

The back propagation algorithm is used in layered feed-forward ANNs. This means that the artificial neurons

are organized in layers, and send their signals ―forward‖,

and then the errors are propagated backwards. The network receives inputs by neurons in the input layer, and the output of the network is given by the neurons on an output layer. There may be one or more intermediate hidden layers. The back propagation algorithm uses supervised learning, which means that we provide the algorithm with examples of the inputs and outputs we want the network to compute, and then the error (difference between actual and expected results) is calculated. The idea of the back propagation algorithm is to reduce this error, until the ANN learns the training data. The training begins with random weights, and the goal is to adjust them so that the error will be minimal.

IV. CONCLUSION

The reconstruction of 3D objects in 3D graphics systems using neural networks is an interesting subject for research. The applications of neural network representation are many, such as: Computer Vision and Error Detection based on neural network matching, Generation of 3D data for computer aided design, Databases of 3D objects for fast querying and finally 3D reconstruction. The goal of this paper is to prove that the process of reconstruction and representation of 3D object can depend completely on multilayer feed-forward neural network. The results show that object reconstructed by neural network is more accurate than object generated by 3rd order polynomial, while object represented by neural network is more exact, accurate and compact at the same time. Thus, the goal of this paper is achieved and this will be a basic step for a series of new ideas that should be explored in future.

V. SIMULATION RESULTS

3D Viewer Screenshot:

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques in processing an image. There

X-SLICE VIEW:

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques in processing an image. There

Y-SLICE VIEW:

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques in processing an image. There

37

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques in processing an image. There

Journal of Computer Applications (JCA) ISSN: 0974-1925, Volume VI, Issue 2, 2013

Z-SLICE VIEW:

Journal of Computer Applications (JCA) ISSN: 0974-1925, Volume VI, Issue 2, 2013 Z-SLICE VIEW: Performance Analysis:

Performance Analysis:

   
 

125 sec

Feed Forward Back Propagation NN
Feed Forward Back
Propagation NN
15 sec
15 sec

REFERNCES

[1] K Susheel Kumar, Vijay Bhaskar Semwal, and Shitala Prasad, ―Generating 3D Model Using 2D Images of an Objectin International Journal of Engineering Science and Technology, 2011. Vol 3. pp 406-415

[2] S. M. Seitz and C. R. Dyer, ―Photorealistic Scene Reconstruction by Voxel Coloring,International Journal of Computer Vision, Volume 35, Issue 2, pp 151173, June 1999 [3] Culbertson, T. Malzbender and G. Slabaugh, ―Generalized Voxel Coloring,In Proceedings of the ICCV Workshop, Vision Algorithms Theory and Practice, Springer-Verlag Lecture Notes in Computer Science 1883, pp 100-115, 1999. [4] O. Ozun, U. Yilmaz and V. Atalay, ―Comparison of Photo consistency Measures Used in Voxel Coloring,Msc. Thesis, Computer Vision & Remote Sensing, Berlin University of Technology, Germany,

2002

[5]

M.

R.

Stevens,

B.

Culbertson

and T. Malzbender, ―A

Histogram-based Color Consistency Test for Voxel Coloring,In

Proceedings of International Conference on Pattern Recognition, Volume 4, pp. 118- 121, 2002

[6] O. Batchelor, R. Mukundan, R. Green ―Incremental Voxel Coloring by Ray Traversalin Computer Graphics, Imaging and Visualization,

2006

International Conference on Issue Date : 26-28 July 2006 On page(s):

396 - 401

[7] M. Heath, S. Sarkar, T. Sanocki and K. Bowyer, ―Comparison of edge detectors: a methodology and initial studyin Computer Vision and Pattern Recognition, 1996. Proceedings CVPR '96, 1996 IEEE Computer Society Conference on Issue Date: 18-20 Jun 1996 On page(s): 143 14 [8] V. Chhabra, 2001. Reconstructing specular objects with image based rendering using color caching. Master’s thesis, Worcester Polytechnic Institute

[9] Dianyong Zhang, Zhenjiang Miao; ―Photorealistic 3d Volumetric

Model Reconstruction By Voxel Coloringin Paparoditis N., Pierrot-Deseilligny M., Mallet C., Tournaire O. (Eds), IAPRS, Vol. XXXVIII, Part 3B Saint-Mandé, France, September 1-3, 2010 [10] Broadhurst, R. Cipolla; ―A Statistical Consistency Check for the Space Carving AlgorithmDept. of Engineering, Univ. of Cambridge, Cambridge, CB2 1PZ

38

[11] Frank Porter, ― Testing consistency of two histograms, California Institute of Technology, Lauritsen Laboratory for High Energy Physics, March 7, 2008

[12] Patrick de Oude, ―Two color consistency checks for voxel

coloring, June 24, 2004 [13] Dr. Kalpana Yadav , Ms. Ankita, Megha Pahwa, Niharika Sachdeva, Shruti Choudhary, ―Photorealistic 2D to 3D Reconstruction using Improved GVC-LDIin International Journal of Computer Technology and Applications, Vol. 3, Issue 2, April,2012. [14] J.D. Foley, A. van Dam, S.K. Feiner and J.F. Hughes. Computer Graphics: Principles and Practice. Addison-Wesly Publishing Co, section 12.6, 1990. [15] Call for Proposals on 3D Video Coding Technology, Standard ISO/IEC JTC1/SC29/WG11, Mar. 2011.

[16] Joint Call for Proposals for Video Coding Technology, Standard ISO/IEC JTC1/SC29/WG11, Jan. 2010. [17] Report of Subjective Test Results from the Call for Proposals on 3D

Video Coding, Standard ISO/IEC JTC1/SC29/WG11, Nov. 2011. [18] Text of ISO/IEC DIS 23008-2 High Efficiency Video Coding, Standard ISO/IEC JTC1/SC29/WG11, July 2012. [19] The Virtual Reality Modeling Language, Standard ISO/IEC DIS 14772-1, Apr. 1997.

[20] Advanced Video Coding for Generic Audiovisual Services, Standard ISO/IEC JTC 1, Mar. 2012. [21] P. Kauff, N. Atzpadin, C. Fehn, M. Müller, O. Schreer, A. Smolic, and R. Tanger, ―Depth map creation and image based rendering for advanced 3DTV services providing interoperability and scalability,‖ Signal Process., Image Commun., Special Issue 3DTV, vol. 22, no. 2, pp. 217234, Feb.

2007.

[22] S.-Y. Kim and Y.-S. Ho, ―Mesh-based depth coding for 3D video using hierarchical decomposition of depth maps,‖ in Proc. IEEE Int. Conf. Image Process., San Antonio, TX, USA, Sep. 2007, pp. V-117V-120. [23] W.-S. Kim, A. Ortega, P. Lai, D. Tian, and C. Gomila, ―Depth map coding with distortion estimation of rendered view,‖ Proc. SPIE, vol. 7543, pp. 75430B-175430B-10, Jan. 2010.

[24] J. Konieczny and M. Domanski, ―Depth-based inter-view prediction of motion vectors for improved multiview video coding,‖ in Proc. IEEE True Vis., Capture, Transmiss. Display 3D Video, Tampere, Finland, Jun. 2010, pp. 14. [25] J. Konrad and M. Halle, ―3-D displays and signal processing,‖ IEEE Signal Process. Mag., vol. 24, no. 6, pp. 97111, Nov. 2007

[26] H.-S. Koo, Y.-J. Jeon, and B.-M. Jeon, ―Motion skip mode for MVC,‖ ITU-T and ISO/IEC JTC1, Hangzhou, China, Tech. Rep. JVT-U091, Oct.

2006.

[27] M. Lang, A. Hornung, O. Wang, S. Poulakos, A. Smolic, and M. Gross, ―Nonlinear disparity mapping for stereoscopic 3D,‖ ACM Trans. Graph., vol. 29, no. 3, pp. 75:175:10, Jul. 2010. [28] E.-K. Lee, Y.-K. Jung, and Y.-S. Ho, ―3-D video generation using foreground separation and disocclusion detection,‖ in Proc. IEEE True Vis., Capture, Transmiss. Display 3D Video, Tampere, Finland, Jun. 2010, pp. 14. [29] Y. Liu, Q. Huang, S. Ma, D. Zhao, and W. Gao, ―Joint video/depth rate allocation for 3D video coding based on view synthesis distortion

model,‖ Signal Process., Image Commun., vol. 24, no. 8, pp. 666681, Aug. 2009.

[30] M. Maitre and M. N. Do, ―Shape-adaptive wavelet encoding of depth maps,‖ in Proc. Picture Coding Symp., Chicago, IL, USA, May 2009, pp.

14.

[31] P. Merkle, C. Bartnik, K. Müller, D. Marpe, and T. Wiegand, ―3D

video: Depth coding based on inter-component prediction of block partitions,‖ in Proc. Picture Coding Symp., Krakow, Poland, May 2012, pp. 149152.

[32] P. Merkle, Y. Morvan, A. Smolic, D. Farin, K. Müller, P. H. N. de

With, and T. Wiegand, ―The effects of multiview depth video compression on multiview rendering,‖ Signal Process., Image Commun., vol. 24, nos. 12, pp. 7388, Jan. 2009. [33] P. Merkle, A. Smolic, K. Müller, and T. Wiegand, ―Efficient prediction structures for multiview video coding,‖ IEEE Trans. Circuits Syst. Video Technol., vol. 17, no. 11, pp. 14611473, Nov. 2007. [34] Y. Morvan, D. Farin, and P. H. N. de With, ―Platelet-based coding of depth maps for the transmission of multiview images,‖ Proc. SPIE, vol. 6055, p. 60550K, Jan. 2006.

[35] K. Müller, P. Merkle, and T. Wiegand, ―3D video representation using depth maps,‖ Proc. IEEE, Special Issue 3D Media Displays, vol. 99, no. 4, pp. 643656, Apr. 2011.

Journal of Computer Applications (JCA) ISSN: 0974-1925, Volume VI, Issue 2, 2013 Z-SLICE VIEW: Performance Analysis:

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques

BIOGRAPHY

B.Ramesh received the PG degree in Applied Electronics and UG degree in Electronics and Communication Engineering from the Anna University of Chennai, Chennai, India, in 2013 and 2011, respectively. He has been with the PMC Tech Institute for Electronics, Hosur, since 2013, where he is currently the Assistant Professor of the Electronics and Communication Engineering, Electronics Department. His current research interests include representation, coding and reconstruction of 3-D scenes in free viewpoint video scenarios and coding, multiview applications, and combined 2-D/3-D similarity analysis. He has been involved in international standardization activities.

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques BIOGRAPHY B.Ramesh received the PG
Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques BIOGRAPHY B.Ramesh received the PG

R.Kannan received the PG degree in VLSI Design and UG degree in Electronics and Communication Engineering from the Anna University of Chennai, Chennai, India, in 2013 and 2010, respectively. He has been with the PMC Tech Institute for Electronics, Hosur, since 2013, where he is currently the Assistant Professor of the Electronics and Communication Engineering, Electronics Department. His current research interests include representation, coding and reconstruction of 3-D scenes in free viewpoint video scenarios and coding, multiview applications, and combined 2-D/3-D similarity analysis. He has been involved in international standardization activities.

39

Three-Dimensional Face Reconstruction from a Single Image using Soft Computing Techniques BIOGRAPHY B.Ramesh received the PG