Stereoscopic conversion of monoscopic video by the transformation of vertical-to-horizontal disparity

摘要: We present a new method for converting monoscopic video to stereoscopic video. The key characteristic of our proposed is that it can process non-horizontal camera/object motion existing in most image scenes. It well known the causes vertical parallax human eyes and accordingly visual discomfort. methodology composed four major steps. First, given current frame, we estimate vector each block by conventional matching estimation algorithm. horizontal disparities. Second, norm computed block. Here, disparity eliminated due usage vector. Due unreliability estimated vectors, low-pass filter performed on order enhance reliability. Third, shifted direction vector, which transformed binocular parallax. shift blocks eliminates effects disparity. Finally, all are synthesized then generated. A pair being original its associated produced. With proper 3D viewing devices, users feel depth from seeing image. Preliminary experiments have demonstrated stable pairs be produced applying variety with camera panning and/or object motion.