Mobile-based Colorimeter with a Hybrid Machine Learning Approach for Enhanced Accuracy

This work proposes a cost-effective, user-developed colorimeter leveraging widely available mobile devices such as smartphones, tablets, and embedded electronic boards with programming and wireless communication capabilities. To achieve high accuracy, these devices utilize a specific color model developed through a combination of hybrid color correction and Random Forest regression. This approach enables precise estimation of color values from camera-measured RGB signals. Experimental proof of concept on four smartphones demonstrates promising average ΔE00 of < 0.35 and ΔE00max of < 7.5 which are acceptable for general-purpose applications. It also confirms that the color model generated by the proposed approach helps accurately estimate color values from camera-measured RGB signals. Future potential uses include cloud-based colorimeters, facilitating the sharing of color models and data for large-scale collaborative experiments.