System and method producing high definition video from low definition video

摘要

Provided is system and a method of producing a high definition video from a low definition video. According to one or more embodiments, the method performs a trajectory estimation and a matching operation only with respect to a feature region pixel of a previous frame of the low definition video, which is different from all pixels of the previous frame, thereby reducing an amount of computation to improve efficiency of a memory, producing the high definition video in real time, and producing a more accurate high definition video with respect to a feature region.

主权项

1. A method of producing a high definition video from a low definition video, the method comprising:
individually classifying pixels or blocks of a current frame of the low definition video into a feature region and a uniform region in accordance with a respective spatial frequency of each pixel or block, with pixels or blocks classified into the feature region respectively being feature region pixels or blocks; and producing a high definition video corresponding to the feature region based on a plurality of frames of the low definition video with respect to the feature region and based on pixel value information from at least one previous frame to the current frame, and producing a high definition video corresponding to the uniform region using interpolation at least between pixels or blocks of the current frame classified into the uniform region, wherein the classifying of the pixels or blocks of pixels of the current frame respectively into the feature region and the uniform region includes classifying a pixel or block of pixels of the current frame having a spatial frequency equal to or greater than a first reference value into the feature region, and classifying a pixel or block of pixels of the current frame having a spatial frequency equal to or less than a second reference value, less than the high spatial frequency, into the uniform region, the producing of the high definition video corresponding to the feature region includes determining pixel values of pixels of a frame of the high definition video corresponding to the feature region by matching feature region pixels of the current frame with feature region pixels of the at least one previous frame determined to correspond to the feature region pixels of the current frame, the determining of the pixel values matches the feature region pixels of the at least one previous frame, based upon a determined difference between the pixel values of the feature region pixels of the current frame and pixel values of corresponding feature region pixels of the at least one previous frame being a third reference value or less, and the third reference value varies depending on whether a feature region pixel of the current frame is determined to be included in a texture region or an edge region.