Inpainted image quality assessment based on machine learning

In many cases inpainting methods introduce a blur in sharp transitions in image and image contours in the recovery of large areas with missing pixels and often fail to recover curvy boundary edges. Quantitative metrics of inpainting results currently do not exist and researchers use human comparisons to evaluate their methodologies and techniques. Most objective quality assessment methods rely on a reference image, which is often not available in inpainting applications. This paper focuses on a machine learning approach for noreference visual quality assessment for image inpainting. Our method is based on observation that Local Binary Patterns well describe local structural information of the image. We use a support vector regression learned on human observer images to predict the perceived quality of inpainted images. We demonstrate how our predicted quality value correlates with qualitative opinion in a human observer study.