Super-resolution Reconstruction Algorithm Based On Image Sequence

Super-resolution image reconstruction refers to the technical method to build ahigh-resolution image from existing low-resolution images. Super-resolution imagereconstruction has advantage like non hardware dependency, low cost, there is widerange usage in the military, medical, industrial, public security, transportation, civil andother fields. Therefore, it has important value to application usage.Super-resolution reconstruction of image sequences is procedure to integrate a setof low-resolution images into a high-resolution image by signal processing method.The key technologies and evaluation methods of super-resolution imagereconstruction are discussed. The POCS method is studied, an image sequence-basedmethod of super-resolution image reconstruction in deep, and two important procedures,motion estimation and initial value estimation, which effect POCS reconstruction resultare analyzed especially. A block-matching method is proposed which based doublecross search of initial value estimation. This method uses motion vector correlation toinitial search center prediction, locate large motion vectors with large cross template,and do refinement search to small motion vectors with small cross template. Thismethod reduces the number of search example points, improves motion estimationaccuracy and search efficiency. Tradition POCS algorithm has a short coming whichdeducing picture quality in edge area, this phenomena is improved by estimate theoriginal image in reconstruction procedure by edge detection and gradient interpolation,the improved method is proposed with the edge gradient interpolation. Experimentsshow that, the reconstructed high-solution image improves resolution while keeping agood image detail, to a certain extent, improves the sawtooth effect and imagereconstruction quality.