Site Loader

In recent years face reconstruction is one of the most outstanding research areas in forensic medicine which has the great impact in medical field. The aim of face reconstruction is estimate the unknown face from skull which support in reconstruction and recognition. Face reconstruction is most challenging problem because the accurate face reconstruction is very difficult method, for that reason we propose the craniofacial reconstruction method. Craniofacial remonstration is the application of anthropology and forensic science. Craniofacial reconstruction is reconstruct the face based on shape and surface structure of skin and skull image. In this paper we reconstruct the face from 3D MRI human heads presented. Our proposed work reconstructs the face from 3D image by using the various features. Here we consider features are 3D image skull, face age, gender, Body Mass Index BMI, and race values based on these features we reconstruct the face accurately. Our proposed method analyzes the accuracy and execution time of face reconstruction.

Key terms – craniofacial reconstruction, 3D MRI image, Body mass index, race.
I.INTRODUCTION

Facial reconstruction is the important technique in several research areas, especially in anthropology and forensic medicine. These two fields are widely research the face reconstruction. From now several facial reconstruction techniques are introduced 1, 7, 8 and 9. This paper was proposed the forensic facial approximation technique which is reconstruct the face based on volume and shape of skull. Here facial approximation reconstructs the face by using both scientific standard and artistic skill. These methods classified in various reconstructive methods for reproduce the face. In 2 author proposed the craniofacial reconstruction model which detect the face based on the land mark values on facial points. Here author reconstruct the face to build 3D statistical model of skull and face soft tissues from 3D CT images. The statistical model used for reproduce the face soft tissue from skull. Facial reconstruction is very important in various fields such as accident,
terrorist attack, genocide, large number of death result from wars. Here we implement the three dimensional reconstruction of same skull in each set tissue by the depth measurement. Facial reconstruction by using the computerized techniques which is based on the landmark interpolation procedure. Land mark interpolation constructed by using the single static facial surface template. By using this technique we reconstruct the facial point from skull 4.

Craniofacial reconstruction method reconstructs the face by two types of classifications such as traditional manual method and computer based approach. In traditional manual method face reconstruction consists of physically modeling face on skull replica with clay or pastime. These techniques take long time for facial reconstruction. Computer based approach reconstructs the face in short time 22. In 23 and face reconstructed by using the computerized technique. These types of techniques increase the system flexibility, speed and efficiency. Here reconstructs the face from either 2D or 3D images. Both methods reconstruct the face by using the various features such as Body Mass Index value, stature ethnic group or else race, age and gender. This method face reconstructs is differ from the other methods by the selection of land mark and skull features. Several methods are using the skull template for face reconstruction. Paper 5 was proposes the reconstruction method based on direct anthropometry using calipers which is the standard technique for craniofacial surgical planning and outcome assessment. These methods have the several drawbacks such as time of reconstruction is high, limitation of data, training required and invasiveness. Normally humans are differentiated by the facial features; each person has the different skull structure and soft tissue depth. Here we consider lips, nose, eyes and ears are the facial features. Here we reconstruct the face by these facial features; these features are varied by each person 2. In 6 author proposed face reconstruction by using evaluation of implicit critical social information. For example human face constructs the wide range of systematic information such as age, gender, expressions, race, pose etc., by using these feature we reconstructs the face. These method used in human computer interface and visual surveillance system9.

Contributions

?We reconstruct the face by using the craniofacial reconstruction method which reconstructs the face by the feature extraction.

?We implement the craniofacial reconstruction from the 3D skull and skin image. Here we first find the facial points for each side of image and then we compare the facial points in skull with skin. Thus we reconstruct the face from 3D skull image.

?Our method finally analyses performance of face reconstruction with the parameter of the accuracy and execution time.

Remaining section organized as follow as section 2 describes related work, section 3 illustrates the problem formulation, section 4 detailed proposed system, section 5 we explain the experimental results and we conclude the our proposed work and future work in section 6.
II.RELATED WORK

Several existing methods are involved for face reconstruction such as nonlinear deformable model, manual identification method, computerized three dimensional craniofacial reconstruction technique, and computed tomography. Nonlinear deformable model constructed by author Adel kermi et al 14. Deformable model proposes the computerized 3D MRI facial reconstruction method which is constrained the face by the information of soft tissue thickness in human heads at certain land marks. It is also known as the B- spline free from deformation model because it’s fully based on non-linear registered technique. This model change the soft tissue thickness and locations are changed until reconstruct the accurate face reconstructed. Deformable model reconstruct the face by using the five stages such as generation of simplex meshes, automatic location of land marks, projection of land marks from skin surface to unknown skull, selection of land mark for reconstruction and add anthropometrical constraints and evaluate the deformable model. Here automatic land mark selection and insertion used as the four methods such as, Mean Curvature without uniform selection MEA, Gaussian Curvature without uniform selection GAU, using Mean and Gaussian without uniform selection MEA-GAU and using Mean and Gaussian with uniform selection MEA-GAU-UNI. Paper 15 proposed the manual identification method for facial reconstruction. Here we reproduce the set of real face points by using the comparison of the soft tissue points for each craniometric point conjunction with the skull points. In this method we increase the points until all the anatomical rules are used. Various existing methods are used the craniometric points for facial reconstruction, but it involves the various interpolation method and restriction technique.

Computerized three dimensional craniofacial reconstruction technique which is used for human identification, this technique reconstruct the face without using the any facial template. This technique requires the age, gender and BMI values skull. This approach reconstructs the face from the 3D image of target skull 12. Jimena et al 13 was proposed the tomography tool for facial reconstruction. Computed tomography is the powerful tool for face landmark reconstruction which reconstructs the face by estimation of the morphological variation. These variance calculated by the analyzing of measurement error. Here author identify the three types of land marks which is obtained by using following types, type1- observing the biological structures, that are easy to find successively. Type 2- is the skull geometry which defines as the local maxima and minima curvature. Type 3- defined as the external points for instance end points of breadth. MRI is the faster acquisition method than the manual measurement method. Authors 19 were proposed morphological feature based face reconstruction technique. Here MRI data sets are store the morphological features such as gender, race and age. Based on these features we reconstruct the face. The main problem in face reconstruction of these model is large number of photographs are stored for these characteristics. It requires the large amount of storage space for processing the single image. It leads the high computational complexity. Therefore we choose the head model choose from database based on the features.

III. PROBLEM FORMULATION

Human face identification from 3D MRI skull image is the major issue. Several methods were proposed for craniofacial reconstruction such as manual and statistical model. In manual method the skull points or image are physically move until the accurate face found. The statistical model we reconstruct the face by the depth measurement of soft tissue.

In 17 was proposed computerized craniofacial reconstruction model. This model consists of the craniofacial template that is warped towards the unknown target skull. Here the skull transformed by the PCA based transformation model. In previous works statistical model constructed by the facial shape variations and soft tissue depth measurement. Here we measure the soft tissue depth from the dense of head CT images. The major benefits of these model are requires the sufficiently large amount of data for morphological features such as age, gender, and race. This method has drawback for facial reconstruction, this model use the CT image as input image all CT images are lying down and faces seen in upward direction which leads the low accuracy.

Authors 18 were proposed the facial reconstruction for accurate identification of human face from skeleton. Here we correlate the skull and face points of target image by using canonical correlation analysis which mapping the skull and skin points. This method not like an existing method which mapping the skull and skin points only with the correlation points of statistical model. Here author introduce the region fusion strategy which improves the matching accuracy of the identification method.

Skull statistical shape model was constructed as follows:
S (a) = +(4)
Face statistical shape model was constructed as follows:
F (c) = +(5)
By the designing of above model we meets the several drawbacks such as the high execution time for face reconstruction, lower accuracy results, high uncertainty and high error rate. To overcome the above mentioned problem our proposed work design the novel frame work craniofacial reconstruction which provides the high accuracy and lower execution time for face reconstruction.

IV. PROPOSED WORK

Our proposed work reconstruct the human face in more accurate and faster manner by using the method of craniofacial reconstruction which reconstruct the face from the 3D MRI image.

4.1 3D craniofacial reconstruction

Craniofacial reconstruction aims to estimate an individual’s face appearance from its skull using the information about MRI datasets. MRI data set contains morphological characteristics of human head models such as age, gender, race and BMI. With the development of three dimensional digitalization technologies, the research on both computers aided and manual sided craniofacial reconstruction has widely received attention. The evaluation of the craniofacial reconstruction has an important significance in improving the craniofacial reconstruction methods. The proposed approach it comprises the following inputs: 3D image of the target skull, 3D image of the skin, set of 72 landmarks placed on the both skull and skin surface where points are inserted by manually. The final input data required by the application is a set of characteristic metrics of the target person: age, gender, race and BMI. Age, gender and race can be deduced from skull morphology that is always known by the user. However, BMI range will be an unknown parameter, which will have to be estimated. Stages of proposed work are land mark insertion module and skin mesh generated module. Land mark module takes a charge of placing each landmark on skull and skin surfaces. In this phase 72 landmarks are inserted in both skull and skin surfaces placed in all views such anterior view and lateral views. Based on this information, skin mesh is generated.

Fig.1. Racial groups of human faces in various countries Indian, South Asian, Caucasian, and African American

4.2 Land mark Insertion Module

Manual landmarks insertion was proposed in this approach. For accurate detection of human face, 72 set of landmarks inserted on both skull and skin surfaces in all sides. Set of landmarks inserted on skull surface is shown in fig.2 and the overall architecture for CFR is shown in fig.3.

Landmark insertion module is in charging of placing 72 landmarks on the skull and skin surfaces and gets the input data from MRI data set which contains set of parameters such as age, gender, race and BMI. Here two set of points are used for reconstructing the face. Furthermore, this module focuses on the selection of certain landmarks among the 72 landmarks based on the morphological characteristics. Once these landmarks are selected, we construct the 3D mesh belonging to the skull and skin landmarks. The selected landmarks (skull and skin) coincide with the positions of anatomical points. According to the results from the new desired mesh surface, reconstruction accuracy was evaluated.

Fig. 2. Landmark inserted image

Figure 2 illustrates the location of land mark points in 3D skin and skull image, here figure (a) & (c) shows the side view of skull image, and (b) shows the front land mark points in skull. In figure 1 (d) & (f) shows the land mark points in skin and (e) shows the front land mark points in skin image.

Algorithm 1: 3D-Craniofacial Reconstruction

Input: 3D skull, 3D skin, values from MRI data sets Age, BMI, Race and Gender and set of landmarks for both skull and skin Ci and Si
Output: Reconstructed Face {Rf}

1. Begin
2. Given input image of skull, skin, MRI Data sets and 72 landmarks
3. Landmark inserted for both skull and skin images
4. Build a full 3D Mesh from both set of points
4.1 Consider set of landmarks {Ci & Si}
4.2 for (Ci=1;Ci

Post Author: admin