The automatic color grading of ceramic tiles is still a big challenge, which needs more researching work. The classifying methods based on color means or histograms can't meet current factory requests, because in modern building, various types and color patterns have been applied to floor & wall ceramic tiles. In this paper, a new method has been proposed to tackle the question that is wavelet texture analysis combined with color information. The selection of wavelet has been discussed, as well as the extraction and optimization of feature set based on wavelet decomposition of samples' color images. Knn classifier and the error rate have been used to evaluate classifying performance, and FFFS (Floating Forward Feature Selection) method has been applied to feature-sets' sub-optimization. The experimental results have shown that the sub-optimal wavelet textural features combined with color information well described the samples' clustering based on color, texture and resolution, and the classifying results have taken on very good agreement with experts'. The experiments also have shown the feature optimization greatly decreased feature vectors' dimensionality, and consequently decreased computational complexity.