Interpolation Algorithm And Its Web Testing Platform Based On Image Adaptive Contour Description Model

Image interpolation is a fundamental question of image scaling, demosaicking and super resolution technology, it is also a key technology of image editing. How to avoid zipper effect, reducing the structure and luminance error caused by interpolation, is the key issues of the image interpolation with remaining edge and structure. Furthermore, there is an important application prospect about CCD imaging device for the development and lightweight camera by efficient image interpolation and improving color filter array (CFA) demosaicking quality. This dissertation starts work on adaptive image edge directed interpolation and CFA demosaicking application, the main achievements are as follows:(1)On the basis of researching on the classical image interpolation algorithm, we give an image interpolation method with keeping the features of structure tensor. This method utilizes local structure tensor information of images for capturing image local edge direction, and achieves adaptive structure in interpolation process through directional diffusion. Experiments show that this method can reduce the zipper effect in image interpolation process compared with classical interpolation methods.(2)For the CFA demosaicking problem, we analyze the directional limitations of classic interpolation in demosaicking, and research a demosacking method by local directional interpolation and nonlocal adaptive thresholding. The contrastive experiment results indicate that enhancing the anisotropy and directional adaptability in image interpolation can improve the visual quality of the demosaicking image.(3)We study the mechanism of the local image adaptive contour stencil description model, and give a construction method of the local adaptive contour stencil by using the smallest total variation. On this basis, we propose an improved method of contour stencil and sparse regularization image deblurring orientation interpolation. Experiments show that the improved method can adapt to edges and texture structures of images, and improve the visual quality of images after interpolation.(4)Combined with the method of adaptive contour model and directional interpolation, we give a CFA demosaicking methods. Experiments show that this method achieves better image quality on test sets.(5)Based on Web services, we build an online test software about adaptive contour interpolation and demosaicking, and give performance comparisons of each algorithm and the impact of parameters on performance.