Sharpness Optimization of a 3D Scanner Using Modulation Transfer Function
Duration of the Thesis: 6 months
Completion: December 2017
Supervisors: Dr. Daniel Döring (FARO Scanner Production GmbH), Rene Pfeiffer (FARO Scanner Production GmbH)
Examiner: Prof. Dr.-Ing. Norbert Haala
Image quality assessment is an important consideration for digital imaging systems. The process of image acquisition, processing, compression, storage and reproduction by the imaging systems and mechanical tolerances can cause degradation in quality. Sharpness, resolution, signal to noise ratio are some of the factors considered for judging image quality.
FARO Freestyle3D is a handheld scanning device which computes 3D data of objects with a resolution below 1 mm using photogrammetric methods and structured light. It consists of two infrared cameras in a stereoscopic configuration and a laser for structured beam of light. Due to minimum divergence of the laser, a sharp laser dot pattern is projected on the object. It is essential for the camera to acquire these sharp laser dots on the entire image frame for consequent processing like alignment and registration. Therefore it is important to estimate the image quality produced by the camera. Currently, in the testing phase, the lens position is adjusted manually to get the best focus on the object to acquire a sharp image and a quantitative assessment to measure image sharpness is not available. Hence it involves more human effort and time to test the image quality. If the local variation of sharpness in the image is known, the laser dot density can be increased in the sharper regions which lead to higher accuracy of 3D data.
In the thesis we investigate the local variation of sharpness in an image captured by the infrared camera in FARO Freestyle3D. For this purpose we identify different regions within the image called the Regions of Interest (ROI) to measure sharpness. The ROIs selected are the middle, off-center (diagonal) and four corners regions. Since the camera system has a manual focus, the sharpness of the image will change by changing the lens position. Therefore, we examine the change of sharpness values in the ROIs at different lens positions. Finally, an optimized sharpness value and corresponding lens position for best focus can be determined.
A slanted edge target MTF based approach to estimate sharpness of ROIs as a function of system resolution is used. Modulation Transfer Function (MTF) computes the contrast modulation with respect to spatial frequencies. It basically describes the ability of the imaging system to transfer the object details to an image. The metric MTF50 for comparison of ROI sharpness is used, which correlates to the perceivable sharpness.
A new feature in the Graphical User Interface is developed. With this, the user is able to establish a connection with the infrared camera and capture images in real time to generate MTF plots and compute MTF50 values for the ROIs. A wrapper function using ctypes module is written to extend the camera interface functions into Python. New ROIS i.e Corners and Diagonals are introduced as checkboxes.
Three test scenarios are executed with different sources of illumination (indirect sunlight, residual sunlight, infrared lamp) and the local variation of sharpness is analyzed. The MTF50 data from the tests are plotted with respect to different lens positions and are modeled as Gaussian fit functions. The influence of distortion and high contrast edges on the MTF values is discussed. It is noticed that the sharpness at a given lens position is not the same for all ROIs. With the given camera, the ROIs Corner-1 and Corner-3 have maximum sharpness at a different lens position when compared to the other ROIs. The average range of MTF50 values in the test using IR lamps was higher (0.17 cycles/pixel) than the other two tests (0.14 cycles/pixel). The overall sharpness is obtained from averaging the MTF50 curves of all ROIs. The peak value of the averaged curve is the optimized sharpness value. The lens offset position at this peak value is the optimal lens position. A weighted average is suggested as an improvement, since a simple average is sensitive to outliers.
Burns Peter. Slanted-edge mtf for digital camera and scanner analysis. In Proc. PICS Conf.,IS&T, pages 135–138, 01 2000.
Jackson K. M. Roland. A study of slanted-edge mtf stability and repeatability. Proc.SPIE, 9396, 2015.
Uwe Artmann and Dietmar Wueller. Differences of digital camera resolution metrology to describe noise reduction artifacts. Proc. SPIE, 7529, January 18 2010.