Comparing Image Upscaling Quality Between Photoshop and Clip Studio Paint
Loading...
Presenter(s)
Vera Kilpatrick
Abstract
AI features in the digital art programs Clip Studio Paint and Adobe Photoshop have been increasingly marketed as AI becomes prevalent in all aspects of digital life. Both programs consist of AI upscaling features that they promote. In this paper, a set of images will be created and tested, starting at a base resolution and upscaled to several higher resolutions, as well as downscaled to a few lower resolutions. From the upscaled images, metrics will be qualified based on several aspects of image quality- specifically: sharpness, color preservation, detail preservation, and artifact reduction, all of which will be tested using the Python OpenCV library. The algorithms that will test the images are Histogram to measure similarities in pixel color, Peak Signal to Noise Ratio (PSNR) to measure if any artifacts were created in the scaling, and Structural Similarity Index Measure (SSIM) to measure sharpness of the scaled image. While the program is running, impact on the system will be recorded to compare at the end of the testing phase. The expected outcome is in favor of Adobe Photoshop.
Keywords: Upscaling, downscaling, image quality, AI image processing
College
College of Science & Engineering
Department
Computer Science
Campus
Winona
First Advisor/Mentor
Sudharsan Iyengar
Second Advisor/Mentor
Mingrui Zhang
Third Advisor/Mentor
Trung Nguyen
Start Date
4-24-2025 12:00 AM
End Date
4-24-2025 12:00 AM
Presentation Type
Oral Presentation
Format of Presentation or Performance
Pre-Recorded Video
Closed Captions for Vera Kilpatrick's RCA Day 2025presentation
Comparing Image Upscaling Quality Between Photoshop and Clip Studio Paint
AI features in the digital art programs Clip Studio Paint and Adobe Photoshop have been increasingly marketed as AI becomes prevalent in all aspects of digital life. Both programs consist of AI upscaling features that they promote. In this paper, a set of images will be created and tested, starting at a base resolution and upscaled to several higher resolutions, as well as downscaled to a few lower resolutions. From the upscaled images, metrics will be qualified based on several aspects of image quality- specifically: sharpness, color preservation, detail preservation, and artifact reduction, all of which will be tested using the Python OpenCV library. The algorithms that will test the images are Histogram to measure similarities in pixel color, Peak Signal to Noise Ratio (PSNR) to measure if any artifacts were created in the scaling, and Structural Similarity Index Measure (SSIM) to measure sharpness of the scaled image. While the program is running, impact on the system will be recorded to compare at the end of the testing phase. The expected outcome is in favor of Adobe Photoshop.
Keywords: Upscaling, downscaling, image quality, AI image processing
