| 摘要 |
Proposed is a method for processing digital image data (ID), wherein a step (S2) of motion estimation (ME) is performed based on a subdivision (Sj; j = 0, 1, 2, ...) of images (I1, I2) to be processed. A refinement of the subdivisions (Sj; j = 0, 1, 2, ...) is controlled in the neighbourhood of boundaries of moving objects within said images (I1, I2) based on a global refinement condition. According to the present invention the global refinement condition is based on properties of an already determined motion vector field (MVFj; j = 0, 1, 2, ...). |
