This project presents a new approach to recovering colors from black and white images using the machine learning method and human-based computation process. The project conducts the experiments by implementing a web-based system that collects data for human-based computation and automates colorization and machine learning calculations. The system improves the accuracy by K-means clustering methods and verifies the results using ground truth images. The colored solutions represent the internal color images of objects in human recognition; therefore, they are colored in the most natural colors in human eyes. Then this study discusses the results and points of improvements. In the end, it shows the potentiality to use this combinational method for various image processing problems.