A feasibility analysis of image approximation with image quality assessments

With technological developments, the resolution of the display systems is increasing, causing image data to increase. Several technologies are aiding in optimizing these data during transmission and storage while maintaining quality. Still, the internal data transmission for accessing the memory sub-system during reading and writing cycles consumes significant power while requiring large memory for storage, which is inefficient, especially for portable displays. Hence, this work investigates the feasibility of reducing data through approximation for images in different formats and resolutions for human and computer vision based applications. Different image quality assessment metrics are utilized for performance evaluations with optimum image quality assessment selection according to application requirements. In addition, the feasibility analysis is conducted for three different applications as examples of computer vision. The experimental analysis highlights theoretical and industrial importance by showing that image data can be reduced from 25.0% to 62.5%, while satisfying requirements for HVS-based and computer vision-based applications. This study presents a feasibility analysis of approximation based on image formats and resolutions to improve memory usage efficiency. A comparative study of image quality assessments for image approximation based on mathematical analysis is provided while evaluating approximation limit with image quality assessment selection for human and computer vision uses. Also, the usefulness of image approximation is verified using three example cases for computer vision applications.image