Endoscope Exploration on Knee Surface 15-862 Computed Photography Final Project Chenyu Wu
Motivation • Endoscope is widely used in minimally invasive surgery
Problem • The size and distortion of endoscope images limit its use.
Our Goal • Stitch the endoscope images onto a 3D surface to get a bigger view of the knee surface
Solution • Calibration • Image: Correct the distortion • Camera: Get intrinsic parameters • Map onto surface • Put trackers on the endoscope and bone • Compute the transformation between camera and bone • Texture mapping
Image Calibration • Basic idea – warp! rectify! • Generate a set of checker board and grid images • Define the correspondences between original image and endoscope image • Compute Homograph for each corresponding images • Same endoscope, consider least square to get the best fit homograph
Camera Calibration • Determine intrinsic parameters • Given a set of control points with known x,y,z and u,v • Least Square
Transformation between camera and bone • Determine extrinsic parameters
Texture mapping • Compute the transformation from image to bone (surface) • Given each image, we can find the corresponding vertex and local triangles on the surface • Warp!
References • C. ARLOS. H. ENRIQUE. Q. UARTUCCI. F. ORSTER. C. LÉSIO. L. T. OZZI, “Toward 3d reconstruction of endoscope images using shape from shading”, Proceedings SIBGRAPI 2000, XIII Brazilian Symposium on Computer Graphics and Image Processing, Gramado, RS, pp. 90-96, October, 2000. • C. Soler, M. Cani, A. Angelidis, “Hierarchical pattern mapping”, ACM Trans. On Graphics (TOG), 21(3): 673-680, 2002.