A weighted feature-based image quality assessment framework in real-time

Abstract

Nowadays, social media runs a significant portion of people’s daily lives. Millions of people use social media applications to share photos. The massive volume of images shared on social media presents serious challenges and requires large computational infrastructure to ensure successful data processing. However, an image gets distorted somehow during the processing, transmission, sharing, or from a combination of many factors. So, there is a need to guarantee acceptable delivery content, especially for image processing applications. In this paper, we present a framework developed to process a large number of images in real-time while estimating the image quality. Our quality evaluation is measured based on four methods: Perceptual Coherence Measure, Semantic Coherence Measure, Content-Based Image Retrieval, and Structural Similarity Index. A weighted quality method is then calculated based on the four previous methods while providing a way to optimize the execution latency. Lastly, a set of experiments is conducted to evaluate our proposed approach.