Three-Dimensional Stitching of Binocular Endoscopic Images Based on Feature Points

There are shortcomings of binocular endoscope three-dimensional (3D) reconstruction in the conventional algorithm, such as low accuracy, small field view, and loss scale information. To address these problems, aiming at specific scenes stomach organs, a method 3D endoscopic image stitching based on feature points is proposed. The left right images acquired by moving converting them into point clouds matching. They then preprocessed to compensate for errors caused scene characteristics uneven illumination weak texture. camera pose changes estimated detecting matching adjacent images. Finally, calculated transformation matrix, cloud registration carried out iterative closest (ICP) dense whole gastric organ realized. results show that root mean square error 2.07 mm, view expanded 2.20 times, increasing observation range. Compared with methods, it does not only preserve information but also makes much denser, which convenient doctors measure target areas, lesions, 3D. These improvements will help improve accuracy efficiency diagnosis.