The paper presents a method on the surface defect detection of floor tile based on computer vision. With an improved image subtraction matching and correlation algorithm, the target image and the template image are partitioned into blocks and processed, and then the subarea of the target image is subtracted from the corresponding template image in the data base. Finally the correlation coefficient is used to determine whether the quality of the target image meets the requirements. Experimental results show that this method can accurately achieve the reliable and efficient surface defect detection of floor tile in realtime.This method can detect the 4.5 average defect and the average recognition rate can reach 95.74%,which is better than histogram statistical characteristics and GLCM.So the method is of great practical value of the surface defect detection of floor tile.