Contact at mumbai.academics@gmail.com or 8097636691
Responsive Ads Here

Friday, 2 February 2018

Local Directional Number Pattern for Face Analysis: Face and Expression Recognition(2013)







Local Directional Number Pattern for Face Analysis: Face and Expression Recognition(2013)


This paper proposes a novel local feature descriptor, local directional number pattern (LDN), for face analysis, i.e., face and expression recognition. LDN encodes the directional information of the face’s textures (i.e., the texture’s structure) in a compact way, producing a more discriminative code than current methods. We compute the structure of each micro-pattern with the aid of a compass mask that extracts directional information, and we encode such information using the prominent direction indices (directional numbers) and sign—which allows us to distinguish among similar structural patterns that have different intensity transitions. We divide the face into several regions, and extract the distribution of the LDN features from them. Then, we concatenate these features into a feature vector, and we use it as a face descriptor. We perform several experiments in which our descriptor performs consistently under illumination, noise, expression, and time lapse variations. Moreover, we test our descriptor with different masks to analyze its performance in different face analysis tasks
EXISTING SYSTEM:
In the literature, there are many methods for the holistic class, such as, Eigenfaces and Fisherfaces, which are built on Principal Component Analysis (PCA); the more recent 2D PCA, and Linear Discriminant Analysis are also examples of holistic methods. Although these methods have been studied widely, local descriptors have gained attention because of their robustness to illumination and pose variations. Heiseleet al.showed the validity of the component-based methods, and how they outperform holistic methods. The local-feature methods compute the descriptor from parts of the face, and then gather the information into one descriptor. Among these methods are Local Features Analysis, Gabor features, Elastic Bunch Graph Matching, and Local Binary Pattern (LBP). The last one is an extension of the LBP feature that was originally designed for texture description, applied to face recognition. LBP achieved better performance than previous methods, thus it gained popularity, and was studied extensively. Newer methods tried to overcome the shortcomings of LBP, like Local Ternary Pattern (LTP), and Local Directional Pattern (LDiP). The last method encodes the directional information in the neighborhood, instead of the intensity. Also, Zhanget al. explored the use of higher order local derivatives (LDeP) to produce better results than LBP. Both methods use other information, instead of intensity, to overcome noise and illumination variation problems. However, these methods still suffer in non-monotonic illumination variation, random noise, and changes in pose, age, and expression conditions. Although some methods, like Gradientfaces, have a high discrimination power under illumination variation, they still have low recognition capabilities for expression and age variation conditions. However, some methods explored different features, such as, infrared, near infrared, and phase information, to overcome the illumination problem while maintaining the performance under difficult conditions.
DISADVANTAGES OF EXISTING SYSTEM:
v Both methods use other information, instead of intensity, to overcome noise and illumination variation problems.
v However, these methods still suffer in non-monotonic illumination variation, random noise, and changes in pose, age, and expression conditions.
v Although some methods, like Gradientfaces, have a high discrimination power under illumination variation, they still have low recognition capabilities for expression and age variation conditions.
PROPOSED SYSTEM:
In this paper, we propose a face descriptor, Local Directional Number Pattern (LDN), for robust face recognition that encodes the structural information and the intensity variations of the face’s texture. LDN encodes the structure of a local neighborhood by analyzing its directional information. Consequently, we compute the edge responses in the neighborhood, in eight different directions with a compass mask. Then, from all the directions, we choose the top positive and negative directions to produce a meaningful descriptor for different textures with similar structural patterns. This approach allows us to distinguish intensity changes (e.g., from bright to dark and vice versa) in the texture. Furthermore, our descriptor uses the information of the entire neighborhood, instead of using sparse points for its computation like LBP. Hence, our approach conveys more information into the code, yet it is more compact—as it is six bit long. Moreover, we experiment with different masks and resolutions of the mask to acquire characteristics that may be neglected by just one, and combine them to extend the encoded information. We found that the inclusion of multiple encoding levels produces an improvement in the detection process.
ADVANTAGES OF PROPOSED SYSTEM:
1)    The coding scheme is based on directional numbers, instead of bit strings, which encodes the information of the neighborhood in a more efficient way
2)    The implicit use of sign information, in comparison with previous directional and derivative methods we encode more information in less space, and, at the same time, discriminate more textures; and
3)    The use of gradient information makes the method robust against illumination changes and noise.
SYSTEM ARCHITECTURE:
MODULES:
1.     Face recognition,
2.     Histogram generation,
3.     Expression Recognition,
4.     Face Retrieval


MODULES DESCRIPTION:
1.     Face recognition:
In the first module, we design the system such that first the image dataset folder should be indexed by the user. After index is made, it shows the number of images in the folder which we indexed. Next the query image is selected by the user. The LH and MLH are used during the face recognition process. The objective is to compare the encoded feature vector from one person with all other candidate’s feature vector with the Chi-Square dissimilarity measure. This measure between two feature vectors, F1and F2, of length N is measured. The corresponding face of the feature vector with the lowest measured value indicates the match found.
2.     Histogram generation:
In this module, the histogram is generated based on the query image selected from the image dataset. The horizontal axis of the graph represents the tonal variations, while the vertical axis represents the number of pixels in that particular tone. The left side of the horizontal axis represents the black and dark areas, the middle represents medium grey and the right hand side represents light and pure white areas. The vertical axis represents the size of the area that is captured in each one of these zones. Thus, the histogram for a very dark image will have the majority of its data points on the left side and center of the graph. Conversely, the histogram for a very bright image with few dark areas and/or shadows will have most of its data points on the right side and center of the graph.
3.     Expression Recognition:
We perform the facial expression recognition by using a Support Vector Machine (SVM) to evaluate the performance of the proposed method. SVM is a supervised machine learning technique that implicitly maps the data into a higher dimensional feature space. Consequently, it finds a linear hyperplane, with a maximal margin, to separate the data in different classes in this higher dimensional space. After the histogram identified in the previous module, we extract all the feature automatically and the features are stored separately. Based on the extracted features, the expression is recognized.
4.     Face Retrieval:
In this module, we retrieve the similar images based on the expression recognized on the previous module. The efficiency of the descriptor depends on its representation and the ease of extracting it from the face. Ideally, a good descriptor should have a high variance among classes (between different persons or expressions), but little or no variation within classes (same person or expression in different conditions). These descriptors are used in several areas, such as, facial expression and face recognition.

SYSTEM CONFIGURATION:-

HARDWARE REQUIREMENTS:-


ü Processor             -        Pentium –IV

ü Speed                             -        1.1 Ghz
ü RAM                    -        256 MB(min)
ü Hard Disk            -        20 GB
ü Key Board            -        Standard Windows Keyboard
ü Mouse                  -        Two or Three Button Mouse
ü Monitor                -        SVGA

SOFTWARE REQUIREMENTS:

•         Operating system           : - Windows XP.
•         Coding Language :  C#.Net

No comments:

Post a Comment