Facial Emotion Detection in Low Resolution Image: Review and Challenges

Abstract

Face emotion is a form of non-verbal communication extensively used by human beings. Detecting emotion to ascertain a human's mental state is a vital research area. Emotion detection for images with low resolution is an area where research is going as practical real-world images will be mostly low resolution. Real-world applications like surveillance, forensic and drone imagery capture low-resolution images due to camera resolution, distance of face, and motion of the camera as well as face and may contain noise. Most of time, researchers consider image less than 32* 32 pixels as low-resolution image as there is no fixed definition for categorizing low-resolution images. Human face emotion detection performance depends on factors such as lighting conditions, facial occlusion, and facial movements. The loss of resolution will reduce the image's vital facial details. Performance on facial emotion recognition is improving with newer deep learning models, but model accuracy remains a major issue.

This paper will elaborate methods and algorithms used in face emotion detection in low resolution conditions. The paper will summarize different datasets, methodology and performance parameters and challenges.