The Contour Completion Algorithm Research Based On Partially Occluded Objects

In2.1D sketch, the occlusion relation is widespread, an object may be occluded byother objects and itself, if occlusion extent is different, so the loss of information of theoccluded objects are also different.2.1D sketch has a concept of the contour completion,that is the recovery of partially occluded object contour. The contour completion on thepartially occluded object can be used for object recognition, motion analysis and objecttracking, this study can also be used for the hidden image, generating stereoscopicgraphics, it is also an important algorithm in graphics rendering.Now, at home and abroad, the contour completion of partially occluded objects hascarried out some research. Most of existing methods for the contour completion, thegenerated contour is different from the original image. Research on contour completion,most of them are recovering the2D contour, and most of them are recovery for smoothcontour, the corner contour completion is less, for the study of3D contour completionis not much. This paper mainly focuses the research on contour completion. Theconcrete research contents and research production are as follows:1. To discuss the present situation of this topic research, The present situationincluding four aspects, that is, T junction detection, occlusion judge, two-dimensionalcontour completion and three-dimensional contour completion. This paper expounds theimage preprocessing. Firstly, do image preprocessing which includes image denoisingand edge contour extraction, and then we detect the T junction, last using the imagesegmentation to extract the object what we need to recovery.2. To propose the method of the corner contour completion based on Euler arcsplines. To summarizes various methods for contour completion, and analyze theadvantages and disadvantages, aim at the current situation of existing corner contourcompletion method is less, the paper puts forward a method of Euler arc splines torecovery the corner contour. The algorithm firstly uses the Levenberg-Marquardt to findthe optimal arc length and curvature, and then separately to calculate the requiredparameters to get the Euler arc splines, finally according to the scalability of the Eulerarc splines to recovery the corner contour, and do experiment with synthetic images andreal images, the results proves that the method is effective and accurate.3. To propose the contour completion of the3D object in the3D scene. Due to theless of the3D contour completion, we will study it. We first introduce some knowledge of3D contour completion, occlusion removal and3D object model. And in the basis ofthe occlusion removal and3D object model, to introduce the3D Euler spiral, so usingit to recovery the missing contour of the three-dimensional object model.