Intelligent Model for Facial Image Quality Enhancement

Abstract

Image is an important artefact in many real world applications. Image improves human perception in understanding. In fact, images play pivotal role in the contemporary era and give higher importance for its storage and retrieval in the context of modern technologies like cloud computing and innovations in image processing. In images, human face is very important for recognition or identification. Images with human face are known as facial images. Such images can participate in biometric authentication, security and forensics. In fact, facial images can be used to uniquely identify humans across the globe. Therefore, facial images are given paramount important in different applications. Having recognized the wide usage of facial images, it is also important to mention issues associated with mediocre quality of facial images. When quality of facial images is low, it affects the performance of application that uses such images for intended purpose. The reasons for low quality of images are many. The image might be captured with different pose, lighting conditions and so on. When the level of quality of facial image is inadequate for processing, the image processing applications will produce less accurate results. Therefore, enhancing facial image quality prior to using it in real world applications is a challenging problem that needs to be addressed. Many solutions came into existence as found in the literature. However, with respect to facial images, it is understood that there is scope of further research and enhancements. In this project, a new methodology is proposed for enhancing quality of facial images. As there is wide usage of facial images in the real world applications, this research has assumed significance. The contributions of this research are as follows. First, present state of the art on techniques to enhance facial images is investigated and reviewed. Second, a framework or Intelligent model in other words is designed for facial image quality enhancement. Third, an algorithm named Variance and Sharpness Based Facial Image Quality Enhancement (VS-FIQE). It takes facial image, Gaussian function’s sigma value and variance threshold as input and produces enhanced quality image. A prototype application is built to demonstrate proof of the concept. The empirical study revealed that the proposed methodology is useful in improve quality of facial images. This prototype can be enhanced further and integrated with decision support systems in real time.