Analysis of Digital Image Quality Between Conventional Design Outputs and Generative Artificial Intelligence Images Using Digital Image Processing Methods

Abstract

Recent developments in Artificial Intelligence (AI) have introduced innovative approaches to digital image production, enabling automated image generation from text-based instructions. The increasing adoption of AI image generation tools has created a need for objective methods to evaluate the quality of their outputs in comparison with images produced through conventional design techniques. This research investigates the visual quality differences between AI-generated images and images created using traditional graphic design software. An experimental quantitative method was applied by analyzing image pairs that shared identical resolutions and file formats. Image quality evaluation was performed using the Mean Squared Error (MSE) metric, which quantifies pixel-level discrepancies between two images. The experimental results demonstrate that MSE can effectively identify variations in image characteristics and similarity. Lower MSE values indicate a closer correspondence between compared images, whereas higher values reflect more substantial deviations in pixel composition. The findings highlight the applicability of MSE as a quantitative tool for image quality assessment and provide additional insight into the performance of AI-based image generation technologies within digital image processing applications.