METHOD FOR REPRODUCING AND RECONSTRUCTING SUPER-RESOLUTION IMAGE FROM COMPRESSED LOW-RESOLUTION IMAGES

摘要

PURPOSE: A method for reproducing and reconstructing a super-resolution image from a compressed low-resolution image is provided to reconstruct a high-resolution image from low-resolution images by removing the aliasing effect. CONSTITUTION: A high-resolution image is reproduced from a plurality of low-resolution images that have been discrete-cosine-transformed and quantized. Quantization noise of DCT(Discrete Cosine Transform) coefficient for the low-resolution images is modeled into a discrete parameter having Gaussian distribution. A reference image among the low-resolution images is decided, and then motion parameters among the other images are estimated by obtaining a least square estimation value through Taylor series. A smoothing restraining condition indicating information of the high-resolution image is represented by non-stationary Gaussian distribution and an adaptive smoothing restraining condition is applied to remove compression noise while preserving the contour of the image.