17-09-2014, 09:44 AM
Local Directional Number Pattern for Face Analysis: Face and Expression Recognition
Local Directional.docx (Size: 25.12 KB / Downloads: 7)
ABSTRACT:
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 directionalinformation of the face’s textures (i.e., the texture’s structure) in acompact way, producing a more discriminative code than currentmethods. We compute the structure of each micro-pattern withthe aid of a compass mask that extracts directional information,and we encode such information using the prominent directionindices (directional numbers) and sign—which allows us todistinguish among similar structural patterns that have differentintensity transitions. We divide the face into several regions, andextract the distribution of the LDN features from them. Then,we concatenate these features into a feature vector, and we useit as a face descriptor. We perform several experiments in whichour descriptor performs consistently under illumination, noise,expression, and time lapse variations. Moreover, we test ourdescriptor with different masks to analyze its performance indifferent face analysis tasks
EXISTING SYSTEM:
In the literature, there are many methods for the holisticclass, such as, Eigenfaces and Fisherfaces, whichare built on Principal Component Analysis (PCA); themore recent 2D PCA, and Linear Discriminant Analysis are also examples of holistic methods. Although thesemethods have been studied widely, local descriptors havegained attention because of their robustness to illuminationand pose variations. Heiseleet al.showed the validity of thecomponent-based methods, and how they outperform holisticmethods. The local-feature methods compute the descriptor from parts of the face, and then gather the informationinto one descriptor. Among these methods are Local FeaturesAnalysis, Gabor features, Elastic Bunch GraphMatching, and Local Binary Pattern (LBP). Thelast one is an extension of the LBP feature that was originallydesigned for texture description, applied to face recognition. LBP achieved better performance than previous methods,thus it gained popularity, and was studied extensively. Newermethods tried to overcome the shortcomings of LBP, like LocalTernary Pattern (LTP), and Local Directional Pattern(LDiP). The last method encodes the directionalinformation in the neighborhood, instead of the intensity. Also,Zhanget al. explored the use of higher orderlocal derivatives (LDeP) to produce better results than LBP.Both methods use other information, instead of intensity, toovercome noise and illumination variation problems. However,these methods still suffer in non-monotonic illumination variation, random noise, and changes in pose, age, and expressionconditions. Although some methods, like Gradientfaces,have a high discrimination power under illumination variation,they still have low recognition capabilities for expression andage variation conditions. However, some methods exploreddifferent features, such as, infrared, near infrared,and phase information, to overcome the illuminationproblem while maintaining the performance under difficultconditions.
DISADVANTAGES OF EXISTING SYSTEM:
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.
PROPOSED SYSTEM:
In this paper, we propose a face descriptor, Local Directional Number Pattern (LDN), for robust face recognition thatencodes the structural information and the intensity variationsof the face’s texture. LDN encodes the structure of a localneighborhood by analyzing its directional information. Consequently, we compute the edge responses in the neighborhood,in eight different directions with a compass mask. Then, fromall the directions, we choose the top positive and negativedirections to produce a meaningful descriptor for differenttextures with similar structural patterns. This approach allowsus to distinguish intensity changes (e.g., from bright to dark and vice versa) in the texture. Furthermore, our descriptor uses the informationof the entire neighborhood, instead of using sparse points forits computation like LBP. Hence, our approach conveys moreinformation into the code, yet it is more compact—as it is sixbit long. Moreover, we experiment with different masks andresolutions of the mask to acquire characteristics that may beneglected by just one, and combine them to extend the encodedinformation. We found that the inclusion of multiple encodinglevels 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 neighborhoodin 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 sametime, discriminate more textures; and
3) The use of gradient information makes the method robust against illuminationchanges and noise.
SYSTEM CONFIGURATION:-
HARDWARE REQUIREMENTS:-
Processor - Pentium –IV
Speed - 1.1 Ghz
RAM - 256 MB(min)
Hard Disk - 20 GB
Key Board - Standard Windows Keyboard
Mouse - Two or Three Button Mouse
Monitor - SVGA
SOFTWARE REQUIREMENTS:
• Operating system : - Windows XP.
• Coding Language : C#.Net