An Optimized Deep Learning Approach for Robust Image Quality Classification

Abstract

This study presents a novel methodology for robust classification of image quality, a critical task in the domain of computer vision. The ability to accurately and promptly classify an image as being of inferior quality, due to factors such as lighting, focus, encoding, and compression, is crucial for a wide range of applications, including autonomous vehicles, web search technologies, smartphones, and digital cameras. Moreover, this capability holds significant potential for numerous industrial applications, particularly in the realm of quality assurance in manufacturing processes or outgoing inspections. In response to this requirement, a novel automated system is proposed herein, employing an optimization algorithm to categorize images into six distinct classes: motion blur, white noise, Gaussian blur, poor illumination, JPEG 2000, and high-quality reference images. The proposed framework is evaluated against existing methodologies using a selection of publicly available datasets. Both subjective and objective assessment results will be presented to demonstrate the efficacy of the proposed framework. This work underscores the potential of leveraging optimized deep learning techniques for robust and automatic image quality classification, thereby paving the way for improved quality assurance across diverse industries.