Edge Responsive Local Directional Patterns versus Intensity Dependent Local Patterns: Implementation and Comparative Analysis

Author(s):

Renu S. Patil , Sinhgad College of Engineering, Pune; Vijay B. Baru, Sinhgad College of Engineering, Pune

Keywords:

Local Patterns, Feature Descriptors, Feature Vectors, Expression Recognition, Local Binary Pattern, Local Directional Number Pattern

Abstract:

Local Patterns are an effective way to construct the descriptors of features where local characteristics are considered to be of importance. Such patterns find their use in the task of face recognition and more importantly expression recognition. This paper focuses on the implementation of the intensity variant patterns and the directional pattern for the purpose of expression recognition. The performance of the feature descriptors is compared for the performance analysis. This comparison is carried out on the basis of the recognition time and the recognition efficiency obtained. The system uses both the types of feature extractors one by one and forms the feature vectors from them. These feature vectors are then used to classify the expressions present in the image. The classifier used for the implementation of system is Support Vector Machine. The aim of this paper is to implement the intensity variant local patterns and the edge responsive directional patterns and compare them on the basis of their performance for expression recognition. The conclusion drawn from the comparison is that the edge responsive patterns are more suitable for the feature description than the intensity variant patterns. The patterns that are considered for implementation are Local Binary Pattern and Local Directional Number Pattern. For a broader analysis, the higher order variants such as Ternary and Derivative Pattern are also implemented.

Other Details:

Manuscript Id	:	IJSTEV3I2070
Published in	:	Volume : 3, Issue : 2
Publication Date	:	01/09/2016
Page(s)	:	185-191

Download Article