Motion Estimation Reliability and the Restoration of Degraded Archived Film

The motivation for this thesis has been to improve the robustness of image processing applications to motion estimation failure and in particular applications for the restoration of archived film. The thesis has been divided into two parts. The first part is concerned with the development of an missing data detection algorithm that is robust to Pathological Motion (PM). PM can cause clean image data to be misdiagnosed as missing data. The proposed algorithm uses a probabilistic framework to jointly detect PM and missing data. A five frame window is employed to detect missing data instead of the typical three frame window. This allows the temporally impulsive intensity profile of blotches to be distinguished from the quasi-periodic profile of PM. A second diagnostic for PM is defined on the local motion fields of the five frame window. This follows the observation that Pathological Motion causes the Smooth Local Flow Assumption of motion estimators to be violated. A ground truth comparison with standard missing data detectors shows that the proposed algorithm dramatically reduces the number of falsely detected missing data regions but also results in an increased missed detection rate. The second part is concerned with the applications of Global Motion Estimation. The first contribution of this part is to introduce a robust and precise algorithm for Global Motion Estimation (GME) using motion information from an MPEG-2 stream, which uses a hybrid of two existing GME algorithms. The first algorithm estimates the parameters from the block based motion field obtained from an MPEG-2 stream and also generates a coarse motion-based background segmentation of the frame. The second is a gradient-based technique which estimates the parameters from the image data directly. In this work, an initial estimate for the global motion is obtained using the first approach. The parameters are then refined using the gradient based technique, producing a final global motion estimate in the form of a six-parameter affine model. Unlike existing gradient-based techniques, the motion-based segmentation is used to weight out local motion rather than the size of the image residuals. The performance of the hybrid algorithm is compared to an existing gradient-based technique over a range of challenging sequences and is shown to be more robust to local motion than the existing approaches. The hybrid algorithm is then applied to the problem of mosaicking in sports sequences in which the global motion parameters can be used to make a panorama of an entire shot. The final contribution of the thesis is a new algorithm to segment frames affected by film tear. Film Tear is a form of degradation in archived film and is the physical ripping of the film material. Tear causes displacement of a region of the degraded frame and the loss of image data along the boundary of tear. In [21], a restoration algorithm was proposed to correct the displacement in the frame introduced by the tear by estimating the global motion of the 2 regions either side of the tear. However, the algorithm depended on a user-defined segmentation to divide the frame. This thesis presents a new fully-automated segmentation algorithm which