On the Appearance of Translucent Edges: Supplementary Material

In the supplementary material, we provide additional results and discussion of several parts of the main paper. 1. Effects of Non-Idealities In this section, we show simulations for studying the effects of non-idealities of the wedge geometry on edge radiance profiles. These simulations complement the discussion fo Section 4 of the main paper. Figures 1 and 2 show simulations for the effect of using a non-orthographic camera. We show simulations for two different phase functions, high-variance and isotropic respectively. In each case, we simulate camera angular extents up to 20◦. In all cases, the use of a non-orthographic camera has little effect on the shape of the radiance profiles. Figures 3 and 4 show similar simulations where, instead of the camera, the illumination angular extent is varied. As before, the use of not perfectly collimated light has negligible effect on the shape of the radiance profiles. Figure 5 shows simulations for the effect of surfaces that are not perfectly smooth on radiance profiles. we use a realistic microfacet refraction model [2], and render the edge radiance profile for increasing values of a parameter α controlling the roughness of the surface. A value α = 0 means a perfectly smooth surface, α = 0.1 is relatively rough, and α > 0.4 is very rough surfaces (e.g., etched surface). The simulations shown are for different phase function and geometry configurations. We observe that the radiance profile maintains its characteristic features for a reasonable range of smoothnes. Finally, Figure 6 shows rendered profiles for bevels of increasing radius, as fractions of the field of view. We obposition in te n si ty position in te n s ty position in te n s ty position in te n s it position in te n s ty position in te n s ty Figure 1: Simulations of sensitivity of edge radiance profiles to deviations from an orthographic camera model for a high-variance phase function. The top row compares edge radiance profiles for different anglular extents subtended by camera pixels. The bottom row shows how the edge radiance profile changes within an angle of 20◦ around the corresponding central view direction (shown in black). The average of the profiles in the bottom row produces the profile for shown in the top row for an angle of 20◦. Different columns correspond to different central view directions (left: θv = 25◦, middle: θv = 45◦, right: θv = 65◦). serve that, for small values of the radius, the bevel effectively “erases” all interesting features that are within distance roughly equal to the radius from the geometric edge, while leaving the rest of the profile relatively uneffected. 1.1. Profile Fitting In this section, we describe the algorithm we use in the preprocessing stage of Section 5 of the main paper, to produce analytic fits to rendered edge radiance profiles. We use the database of profiles described in Section 2 of the main paper, and search for a parametric expression that fits all of the profiles in this database, using the insights from position in te n s ty position in te n s ty position in te n s ty position in te n s it y position in te n s it y position in te n s it y Figure 2: Simulations of sensitivity of edge radiance profiles to deviations from an orthographic camera model for an isotropic phase function. The top row compares edge radiance profiles for different anglular extents subtended by camera pixels. The bottom row shows how the edge radiance profile changes within an angle of 20◦ around the corresponding central view direction (shown in black). The average of the profiles in the bottom row produces the profile for shown in the top row for an angle of 20◦. Different columns correspond to different central view directions (left: θv = 25◦, middle: θv = 45◦, right: θv = 65◦).