Whitening level estimation is an application of skin color measurement in the field of cosmetics, which is related to the evaluation of skincare products. Previous skin color measurement methods depend on either contactable devices or cumbersome professional instruments. How to implement non-contactable whitening level estimation in a lightweight manner is urgent in the evaluation of skincare products. In order to tackle this issue, the paper designs a lightweight solution using edge computing to implement whitening level estimation with wearable devices. First, the skin images are collected using camera and represented as Scale-Invariant Feature Transform (SIFT) features; then SIFT feature vectors are input into a trained Reduces Support Vector Ordinal Regression (Reduced SVOR) model to obtain whitening level result. Compared with deep learning models, both SIFT and Reduced SVOR need fewer resources. Thus, the proposed solution can be deployed in edge nodes. The experiments shows the proposed whitening level estimation solution can meet the requirements of mobile devices as well as achieve satisfied mean absolute error (MAE).