Research Of The Image Super-Resolution Reconstruction Algorithm

Image super-resolution reconstruction is a technique to estimate a high-resolution image (or sequence) from a single image or an image sequence having slight difference and combating additive noise and blurring due to the finite detector size and optics. At the same time, it reduces the high cost and conquers the difficulty of enhancing the CCD (charge coupled device) resolution. At present, this technique is widely used for satellite imaging, video monitoring, medical imaging, etc.The development history of image SR construction is looked back, and the popular available algorithms of SR are introduced. Meanwhile, a thorough introduction of the observation model and the components of a typical image SR construction algorithm, including image registration, interpolation and restoration, are given. On the basis of referring to algorithms of other relating technologies, three algorithms are presented.A sequence images super-resolution reconstruction algorithm based on NEDI (new edge-directed interpolation) is proposed. Before interpolation, a frequency domain algorithm by Vandewalle is employed to register a sequence of aliased low-resolution images precisely. A high-resolution image is then obtained using NEDI based on the geometric duality between the low-resolution covariance and the high-resolution covariance.A single image wavelet-based super-resolution reconstruction algorithm based on NEDI is presented. Using NEDI to interpolate the images from applying the inverse wavelet transform to the high frequency subbands are employed to estimate the unknown detail coefficients of higher resolution image.Another single image Contourlet-based super-resolution reconstructionalgorithm based on subbands estimating is proposed. Employ Contourlet transform to provide a fexible multiresolution, local and directional expansion for low-resolution image. Then the inverse wavelet transform with zero-filling the high frequency coefficients is applied to estimate the unknown directional subbands.The algorithms are tested in simulations, and compared to other similar SR methods. For further research, all of them provide precious experience and improving direction.