No Reference Image Quality Assessment for Visual Perception and Accomplishing Tasks using Machine Learning Techniques

Abstract

In the early years of the twenty-first century, researchers have witnessed a proliferation of digital images for representing information. Advancement in multimedia technology is attempting to improve the quality of images in acquiring, processing, storing, and transmission. Images are subject to degradation during the above process, particularly due to compression, which helps in efficient storage and communication. The image quality must be assessed and quantified in each cycle to avoid image degradation. Evaluating the quality of an image is called as Image Quality Assessment (IQA). Quality assessment plays a pivotal role in image processing applications. Some of the applications are: 1) The image acquisition system requires the image quality to be evaluated such that it automatically adjusts the device to capture the high quality image/video. 2) The IQA can act as a evaluation criteria for image processing algorithms (based on the image quality, select the best model over multiple models for the specific task) 3) IQA can be incorporated into image processing models to fine-tune the parameters for optimization. newlineHuman are the end users of the image. Therefore, the obvious way of assessing the image is by looking at its visual perception. Hence human are involved in assessing the images. Involving humans (subjects) in the assessment process is called as subjective IQA. The image quality is quantified with the scores provided by the human, named as the human opinion score. Involving human in the assessment is slow, complicated, expensive, subjective, and challenging to incorporate in automation. Thus resulting in contemporary way of quality assessment which aims to develop an automatic assessment model to quantify or assess the quality of image without involving human. An algorithmic way of evaluating the image is called as objective IQA. newline