Research On The Applications Of Subdivision Modelling In Super Resolution

As a very important field of computer aided geometric design,subdivision schemes shows good properties in the modeling of curves and surfaces.Therefore,this dissertation mainly studies their applications in the field of depth image super-resolution.Firstly,we suppose that an image with the better edge map is of better quality.In view of the good performance of curve subdivision in the smooth curve modeling,we propose to use the curve subdivision model to generate the edge map of the target resolution image.We apply the “node compression method” to overcome the zigzag artifacts in edge.Then,a modified joint bilateral filter is adopted to reconstruct the high resolution depth image with the guidance of the edge map,by which jagged artifacts can be reduced and sharp edges can be preserved well.Secondly,we regard the pixel value as the third dimensional coordinate in space,and view the image as a surface in space.The Doo-Sabin subdivision scheme is used to recover the missing pixels in the image reconstruction process.However,the generated image is blurred at the edge.To optimize the performance of the target image at the edge,the joint trilateral filtering is applied.Final experimental results show that both the boundaries and regions of the image remain smooth.