New Nonphotorealistic Shading Algorithm to Improve the Perceptual Quality of Two-Dimensional Images

We propose a nonphotorealistic shading algorithm to improve the perceptual quality of two-dimensional natural images that do not have depth information. Shading plays an important role in the human perception of three-dimensional shapes; however, an appropriate normal is indispensable to the shading approach. Generally, it is pointed out that luminance variation due to shading is useful as an indicator of shape and that the depth values are strongly related to the luminance distribution of the object. Therefore, first, we compute the surface normals from the variation in the luminance at each pixel. Second, we enhance the shade by controlling the azimuth of the normals on the basis of the angle between each normal and a given light source direction. From experimental results of the natural images and subjective test, we demonstrate that the three-dimensional (3D) appearance can be improved by our shading approach. Moreover, we also propose an advanced shading approach, which is the combination of our shading approach and the process of selecting the suitable light direction, and we show the performance of selecting the suitable light direction and also demonstrate that the 3D appearance can be more enhanced by using the suitable light direction. Last, we investigate the possibilities of the advanced shading approach. (C) 2009 Society for Imaging Science and Technology. [DOI: 10.2352/J.ImagingSci.Technol.2009.53.5.050503]