To further enhance the accuracy of a micro milling process, an effective error measurement method and error compensation method are necessary. An on-machine vision-based error measurement method integrating image re-constructive technology, camera pixel correction, and model comparison algorithm to provide the capability of non-contact measurement for micro-machined workpiece error was developed. With use of Canny Edge Detection algorithm and camera pixel calibration, the edge of the contour of a machined workpiece can be identified and used to re-construct the actual shape of the workpiece. The actual shape will then be compared to the theoretical shape of the workpiece. To prove the feasibility of the proposed methods and system, experiments were conducted. The results have shown the success of enhacing machining accuracy for a micro milling process.