Color‐distance modeling improves differentiation of colors in digital images of hybrid bermudagrass

Abstract

Identifying and reporting color differences is important in turfgrass experimentation. Finding ways to improve color differentiation may be desirable. Research was conducted to determine if color‐distance modeling (∆E*ab) improves differentiation of color in images of hybrid bermudagrass [Cynodon dactylon (L.) Pers. × Cynodon transvaalensis Burtt‐Davy]. Colors differ perceptibly if ∆E*ab >1.0; magnitude of difference increases as ∆E*ab increases. Hue angle (θ) and dark green color index (DGCI) of untreated shoots; shoots treated with petroleum hydrocarbons; and turf treated with N, herbicides, or a growth regulator were determined by digital image analysis. Hue and DGCI means were separated using orthogonal contrasts; ∆E*ab was computed for each contrast. Delta E*ab was a good predictor of color difference magnitude. Association between p‐values for contrasts of θ or DGCI and ∆E*ab were strong; smaller p‐values corresponded to larger ∆E*ab. When log10 p‐values for θ contrasts for ‘Tifway’ turf treated with N, herbicides, and a growth regulator were regressed on ∆E*ab, r2 = .9657 (y = − 0.602 − 0.9512x); for DGCI, r2 = .8743 (y = − 1.6097 − 0.8743x). Computing ∆E*ab helped avoid making false conclusions. Neither θ nor DGCI differed for Tifway treated with mesotrione or glyphosate (p = .2049, .6195) but ∆E*ab was 3.8 due to differences in saturation (p = .0137) and brightness (p = .0067). Computing ∆E*ab improved differentiation of turfgrass color by enhancing the accuracy of difference detection. Improving the ability to differentiate colors and avoidance of making false conclusions has advanced reporting of turfgrass color data.

Predicting Munsell color for turfgrass leaves

“…Color error between leaf blades of Champion and TifEagle was noticeable with ∆ E * ab = 6.2 (Table 7). Values of ∆ E * ab > 3.5 show clear differences between colors (Berndt et al., 2020). Compared to TifEagle the color of Champion appeared brighter (∆ L * = 79.70 − 82.49 = −2.79) with less yellow (∆ b * = 26.37 − 21.09 = 5.28), less saturation (∆ C * = 36.59 − 31.82 = 4.77), and a larger hue angle (∆ h° = 133.87 − 138.49 = −4.62), resulting in a larger value for dark green color index (∆DGCI = 0.532 − 0.581 = −0.049).…”

“…The 16‐bit sRGB TIFF images were opened with Adobe Photoshop (PS, Release 22.4.2, Adobe Systems, Inc., San Jose, CA) and re‐saved in PS 15‐bit+ L * a * b * mode. This digital file arrangement allowed for working directly in L * a * b * color space, where L * a * b * is a device‐independent three‐dimensional opponent color model (Berndt et al., 2020). The L * references perceptual lightness or brightness.…”

“…The Euclidian distance (error) between cultivar colors in L * a * b * was calculated using Equation (). $$$$\begin{equation}\Delta E{^*}_{ab} = {\left[ {{{\left( {\Delta L^*} \right)}}^2 + {{\left( {\Delta a^*} \right)}}^2 + {{\left( {\Delta b^*} \right)}}^2} \right]}^{1/2}.\end{equation}$$$$Values >3.5 indicated perceived color differences (Berndt et al., 2020). Color error was also calculated in terms of L * C * h * color space, where L * is lightness (∆ L *), C * is chroma (∆ C *), and h * is hue (∆ h *).…”

“…Values >3.5 indicated perceived color differences (Berndt et al, 2020). Color error was also calculated in terms of L*C*h* color space, where L* is lightness (∆L*), C* is chroma (∆C*), and h* is hue (∆h*).…”

“…Objective color measurements may mitigate issues with subjective color matching. Quantitative color measurements yield color metrics including hue, saturation, and brightness, CIE‐ L * a * b * color space (CIE, International Commission on Illumination, 2006, 2015, Vienna AT, CIELAB Colour) and CIE‐ xyY color space ( xyY ), and indices such as dark green color index or normalized difference vegetation index (Berndt et al., 2020; Karcher & Richardson, 2003; Karcher & Richardson, 2005; Kerr, 2010; Zhang et al., 2017). Programs specific to turfgrass linking color to metrics or indices may be helpful in reducing subjectivity (Turf Analyzer) (Karcher & Richardson, 2005).…”