A Psychophysical Evaluation of Texture Compression Masking Effects.

Area of Science:

• Computer Graphics
• Perceptual Computing
• Image Processing

Background:

• Lossy texture compression is vital for reducing GPU memory and bandwidth.
• Traditional quality metrics like Peak Signal-to-Noise Ratio (PSNR) are insufficient for evaluating compressed textures due to perceptual masking effects.
• Masking effects occur when textures are rendered on surfaces, interacting with lighting and other texture maps.

Purpose of the Study:

• To conduct the first psychophysical experiment on the perceptual impact of texture compression artifacts on rendered images.
• To investigate how compression bit rate, light direction, and texture map content (diffuse and normal) influence perceived artifact visibility.
• To evaluate the effectiveness of existing image quality metrics in predicting perceived compression artifacts.

Main Methods:

• A psychophysical experiment was designed to collect human perception data on rendered textured images.
• Various compression bit rates, lighting conditions, and texture map types (diffuse, normal) were systematically varied.
• Collected data were used to analyze masking effects and compare the performance of different image quality metrics.

Main Results:

• Significant masking effects were observed between diffuse and normal map compression artifacts.
• Peak Signal-to-Noise Ratio (PSNR) applied to individual textures poorly predicted the visual impact of compression artifacts in rendered scenes.
• Analysis revealed the limitations of standard image quality metrics in accurately predicting perceived artifact visibility.

Conclusions:

• Perceptual masking is a critical factor in texture compression quality assessment.
• Relying solely on PSNR for individual texture maps is inadequate for evaluating real-world rendering quality.
• Recommendations for improved texture compression quality evaluation and a practical application for approximating distortion on rendered 3D shapes were provided.