I. INTRODUCTION:

Advancement in Biometrics advances the law enforcement agencies by providing a means of identifying criminals with lesser amount of time. Visual Biometric along with Fingerprint recognition which uses the ridges and valleys found on the surface tips of a human finger to determine identity of criminals. It also provides Face recognition which gives the analysis of facial features for the recognition of an individual’s identity.

In many cases the facial photograph of a suspect is not available. In these circumstances, drawing a sketch following the description provided by an eyewitness or the victim is a commonly used method to assist the police to identify possible suspects. To draw forensic sketches police artist requires a large amount of training in drawing and sculpting. Composite sketches instead requires several hours of training which allows even non artist to compose a sketch with the help of composite sketch software, becomes a perfect alternative to provide assistant in investigation. Using composite sketches is advantageous as it is less time consuming and more economic. Some of the most widely used facial composite software kits include EvoFIT, Photo-Fit, FACES, Mac-a-Mug and IdentiKit. Some of the most widely used facial composite software kits include EvoFIT, Photo-Fit, FACES, Mac-a-Mug and IdentiKit. Some are used to detect Primary Facial Components such as Hair, Eyebrow, eyes, nose, mouth, shape, eyeglasses, etc. and secondary Facial Components such as Smile lines, moles, scar and tattoo, etc. The EvoFIT is an evolutionary facial composite system, which is different from component based systems in that it creates a holistic likeness to the suspect using a genetic algorithm based on several selected shapes and textures that most resemble a suspect [1]. Figure 1 shows the procedure for creating a composite using FACES. In this facial composite system, each facial component is selected from a candidate list shown on the right side of the system GUI.

The difference between forensic sketches (hand drawn sketches) and computer generated composite sketches can be seen in Fig. 2. Compared to face photos (mug-shot), both hand drawn sketches and composite sketches lack detailed texture, especially around the forehead and cheeks. However, artists can depict each facial component with an exact shape, and even shading. Thus, artist-drawn sketches can usually capture the most distinctive characteristics of different faces. By contrast, facial components in a composite sketch must be approximated by the most similar component available in the composite software’s database. Moreover, the psychological mechanism of an artist guarantees that hand drawn sketches look natural while composite sketches may look synthetic. Therefore, while we may easily recognize a person from his hand-drawn sketch, it is often more challenging to identify a person from a composite sketch. Similar observations have also been reported in the cognitive psychology community, a survey showed that 80% of the officers in law enforcement agencies used computer generated composites [2].

Fig. 1.Procedure for creating a composite sketch using FACES [4].

Despite this high percentage of law enforcement agencies using computer generated composites, the application of automated face recognition algorithms to computer generated composites has not been adequately studied [3]. By contrast, sketch recognition, both viewed sketches and forensic sketches, has received relatively more attention. In this paper, we present a study of a face recognition system to match computer generated facial composites to facial photographs or mugshots. The design objectives of the proposed work are to (i) provide a common representation for both composite sketches and face photos that can diminish intrapersonal variations while still maintaining interpersonal discriminability, (ii) leverage the component-based approach by which computer generated composites are formed, and (iii) effectively match composite sketches against large-scale mugshot gallery databases [1].

II. RELATED WORK:

Research on sketch matching started only since a decennium. Because of the inaccessibility of standard public database for forensic sketches, most of the research is done on viewed sketches only for last 10 years. On viewed sketches, most of the early work is done by Tang et al. [5]-[7]. A Synthetic photograph is generated from the sketch in these works; and then matching is performed with standard face recognition algorithms. Only one paper is published in forensic sketch matching till date. Klare and Jain [8] published a Local Feature based Discriminant Analysis (LFDA) approach for matching forensic sketches to mug shot photos.

Fig.2. Difference between hand drawn sketches and composite sketches (a) Face photographs, (b) the corresponding viewed sketches, (c) the corresponding computer generated composite sketches synthesized using FACEs [4]. Both the viewed sketches and composite sketches were constructed while viewing the face

There are lots of problems in forensic sketch recognition compared to normal face recognition in which both probe and gallery images are photographs. The fineness of sketches whether they may be viewed sketches or forensic are different from large mug-shot gallery. Most of the work done previously is principally focused on forensic sketches or viewed sketches. Forensic sketches have additional problems compared to viewed sketches. Due to the fractious nature of the memory, the exact visual aspect of the criminal cannot be remembered by the spectator. This leads to an incomplete and inaccurate depiction of the sketches which reduces the recognition performance considerably.

To handle such difficulties, Klare et al. developed a local feature-based discriminant analysis (LFDA), which learns a discriminative representation from partitioned vectors of SIFT [9] and LBP [10] features using multiple discriminative subspace projections. Component-based face recognition methods (which are used in this study) were studied in [11]-[15]; however, these algorithms either directly utilized intensity features that are sensitive to changes in facial appearance or employed supervised algorithms for classification whose performance is sensitive to the amount of training data available. Moreover, these algorithms were mainly proposed to resolve the misalignment problem in photo-to-photo face matching, and do not address the heterogeneous modality gap present when matching computer generated composite sketches to facial photographs.

III. COMPONENT EXTRACTION AND DISTANCE MEASUREMENT:

In order to extract facial component and measure the distance between these components following three steps are performed.

1) Face normalization using a geometric transformation and color space conversion.

2) Facial component localization using Viola-John algorithm, followed by ASM.

3) Per component feature extraction using nearest neighbor symmetry and matrix normalize cross-correlation followed by computing length between facial constituent.

A. Preprocessing of mugshots

A novel preprocessing technique is discussed in this section. This preprocessing is different from the conventional face recognition techniques where the face is preprocessed so that region from forehead to chin and cheek- cheek is visible. Here we proposed the algorithm so that the hairline and neck region along with ears are also visible.

Fig 3. Preprocessed image done with our proposed method.

The external features of the face are not lost in the preprocessed image.

B. Facial component localization

It’s a true challenge to build an automated system which equals human ability to detect faces and estimates human body dimensions from an image or a video. The conceptual and intellectual challenges of such a problem, because faces are non-rigid and have a high degree of variability in size, shape, color and texture. It’s a commonly known fact that human face is symmetric about vertical axis. From both neurological and computational point of view symmetry plays an important role in facial part dimensions as it has been demonstrated that an exceptional dimension reduction can be made by taking into account facial symmetry [16].

The present method usually returns the image location of a rectangular bounding box containing a face. Also Active shape modeling [17] is applied to image to get landmarks on the face. About 20 landmarks are detected. Figure (4) shows image with facial component enclosed within rectangular box and face annoted with 20 landmarks.

Fig. 4. Facial element localization (a) input image, (b) Facial elements detection, (c) Facial landmark recognition using ASM.

C. Feature extraction

We extracted total 12 features by using nearest neighbor symmetry [18] and matrix normalize cross-correlation [19], those 12 feature are shown in figure 5 which are as follows: Eye width, Eye height, Nose width, Nose height, Mouth width, Mouth height, Distance between left eye to nose, Distance between right eye to nose, Distance between left eye to mouth, Distance between right eye to mouth, Distance between nose to mouth, Distance between both eyes. Figure shows detected face parts and measured length of facial components, advantage of proposed algorithm is that facial components are detected in spite of pose and scale variation.

Fig. 5: Determination of distances between face components.

IV. EXPERIMENTAL RESULTS:

In this section, we study the performance of the proposed CBR approach for matching composite sketches to facial photographs. A photo-composite database is first built for this experimental study. We analyze the discriminability of the different facial components for composite-to-photo matching. The accuracy of the proposed CBR method is then evaluated using MATLAB Version: 8.1.0.604 (R2013a). Database of 200 mug-shot and composite pairs is use to perform this experiment. Firstly preprocessing of images and composite sketches are done then features are extracted for both followed by score normalization and fusion. Maximum matching scores for composite sketch to mug shot matching is shown in figure 6, which is 99.93% for matched pair.

Table: I -Matching performance of proposed composite sketch-mugshot matching algorithm, CBR using MLBP

Methods

Rank-1 Accuracy

Rank-50 Accuracy

CBR using MLBP

12.2

74.8

CBR using Component lengths Measurement

13.6

76.2

V. CONCLUSION:

This paper investigates a challenging heterogeneous face recognition problem: matching composite sketches to facial photographs. Face varies in rotation, brightness level, size, etc. in different images even for the same person. Those features are independent of face features and will affect the recognition rate significantly. One method to solve the problem we used face normalization using geometric transformation. Face detection is a computer technology that identifies human faces in digital images. It detects human faces which might then be used for recognizing a particular face. We detect the face by using Viola–Jones object detection technique. After that we extract the 12 feature component of detected face it is may be composite sketches or mug shot. A GUI, provides an intuitive interface between the user and the programming language. A GUI allows the user to bypass MATLAB commands altogether and, instead, to execute programming routines with a simple mouse click or key press therefore we build the GUI.

VI. REFERENCES:

Han H., Klare B. F., Bonnen K., Jain A.K.: Matching Composite Sketches to Face Photos: A Component-Based Approach. IEEE Transactions on Information Forensics and Security, Vol. 8, No. 1, January 2013

Mcquiston D., Topp L., Malpass R.: Use of facial composite systems in US law enforcement agencies. Psychology, Crime and Law, vol. 12, no. 5, pp. 505–517, 2006

Yuen P. C., Man C. H.: Human face image searching system using sketches. IEEE Trans. Syst., Man, Cybern. A, Syst. Humans, vol. 37, no. 4, pp. 493–504, Jul. 2007

FACES 4.0, IQ Biometrix 2011 [Online]. Available: http://www.iqbiometrix.com

X. Wang and X. Tang. Face sketch recognition. IEEE Transactions on Circuits and Systems for Video Technology, 14(1):50–57, January 2004.

Q. Liu, X. Tang H. Jin, H. Lu, and S. Ma. A nonlinear approach for face sketch synthesis and recognition. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, pages 1005–1010, 2005.

X. Wang and X. Tang. Face photo-sketch synthesis and recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 31(11):1955–1967, 2009.

B. Klare, Z. Li, and A.K. Jain. Matching forensic sketches to mug shot photos. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 639–646, March 2011.

Lowe, “Distinctive image features from scale-invariant key points,” Int. J. Computer. Vis., vol. 60, no. 2, pp. 91–110, 2004.

T. Ojala, M. Pietikainen, and T. Maenpaa, “Multiresolution gray-scale and rotation invariant texture classification with local binary patterns,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 24, no. 7, pp. 971–987, Jul. 2002.

B. Heisele, P. Ho, J. Wu, and T. Poggio, “Face recognition: Component- based versus global approaches,” Comput. Vis. Image Understand., vol. 91, no. 1, pp. 6–21, Jul. 2003.

Y. Ivanov, B. Heisele, and T. Serre, “Using component features for face recognition,” in Proc. Automatic Face and Gesture Recognition, 2004, pp. 421–426.

Lin and X. Tang, “Recognize high resolution faces: From macrocosm to microcosm,” in Proc. IEEE Computer Vision and Pattern Recognition, 2006, pp. 1355–1362.

B. Klare, A. Paulino, and A. Jain, “Analysis of facial features in identical twins,” in Proc. Int. Joint Conf. Biometrics, 2011, pp. 1–8.

J. Huang, V. Blanz, and B. Heisele, “Face recognition using component- based SVM classification and morphable models,” in Proc. First Int. Workshop on Pattern Recognition with Support Vector Machines, 2002, pp. 334–341.

P. Viola and M. Jones. Rapid object detection using a boosted cascade of simple features. In Proc. of IEEE Conference on Computer Vision and Pattern Recognition, Kauai, HI, December 2001.

T. Cootes, C. Taylor, D. Cooper, and J. Graham, “Active shape models their training and application,” Computer. Vis. Image Understand., vol. 61, no. 1, pp. 38–59, Jan. 1995.

Yufan Wang, Qiuze Yu, Wenxian Yu, ―An Improved Normalized Cross Correlation algorithm for SAR Image Registration, IEEE IGARSS, 2012.

Vanderbrug, G.J, Rosenfeld, ―Two-Stage Template Matching, IEEE Transactions on Computers, Vol. 60, Issue 11, 1977.

Received on 04.06.2015 Accepted on 22.06.2015

Research J. Engineering and Tech. 6(2): April-June, 2015 page258-262

DOI: 10.5958/2321-581X.2015.00039.2