The use of fractal analysis and photometry to estimate the accuracy of bulbar redness grading scales