Video frames interpolation using adaptive warping

In this dissertation, a strategy for high fidelity scaling of video frames to higher resolutions is introduced. The method employed is a combination of block matching-based motion estimation and optical-based motion estimation, which builds on the Control Grid Interpolation (CGI) methods. Low resolution images are interpolated and used to obtain motion vectors that are then used to warp high definition reference frames. Because motion estimation is an ill-posed problem, robustness is critical in the motion search procedure. To improve the robustness of the derived motion fields, a bidirectional motion estimation method is proposed. A hierarchical motion structure is used to solve the ambiguity in the framework. Improving the spatial quality of video coded at low bit rates is a problem of general interest. Toward this end, a combined forward-backward warping method is proposed to improve the resolution of video encoded with H.264/AVC. The scheme attempts to capture long-term spatial detail by warping high resolution reference frames in accordance with displacement vectors derived from decoded low resolution frames. At low bit rates, the proposed method can achieve better PSNRs and better subjective quality than conventional H.264/AVC for sequences with low to moderate motion. The third application considered in this dissertation is video frame rate up-conversion. A new inter-frame motion compensated interpolation method is proposed that employs motion vector correction based on residual energy. Experimental results show that it can improve the visual quality of the interpolated frames where competing methods fail