Digital Image Stabilization by Adaptive Block Motion Vectors Filtering

Filippo Vella at el introduced a robust algorithm for video sequences stabilization using Adaptive Block Motion Vectors Filtering. According to them Digital Image Stabilization systems can be subdivided in three modules: motion estimation, detection of unwanted movements and compensation. They used proposed algorithm as motion estimator a module that produces block motion vector. Motion Estimation evaluated on blocks gives a motion vector for each block considered. The advantages of this method same motion estimator of the mpeg encoder can be used and a big set of motion vectors can be evaluated for each frame. The drawback is that not all block motion vectors (BMV) are reliable. This algorithm considered two areas as foreground and background and motion estimation is done in these two areas. For each area of the frame, block motion estimation is done and a number of motion vectors are estimated. From them a global motion vector for foreground and one for background is evaluated. To calculate a Global Motion Vector for the areas a single Motion Vector must be evaluated from the set of BMVs. The most frequent vector in the extract as the Global Motion Vector. The square histogram considers the most frequent motion vector in the frame and it will work better than two linear histograms. There two histograms are built for foreground and background and the maximum of each histogram will give a GMV for each region.

When detecting unwanted movements, they considered about GMV in foreground and background. If GMV for foreground and background are equal, the detected motion is considered as the global motion that affects the frame. Otherwise a decision must be taken whether to stabilize accordingly to the first or the second. To decide whether to consider motion of foreground or background the number of blocks that produce the GMV is considered. If the background area GMV is produced by a higher number of blocks than the foreground area GMV, background is stabilized, otherwise foreground is stabilized.

When unwanted motion of the frame is detected, a motion compensation is done. The AGMV is the vector that accumulates the components of the GMV of each frame and to have a stabilization of the whole sequence the Absolute GMV (AGMV) is used. A movement can be classified as jiggling or panning. Discrimination between jiggling and panning has been considered to avoid to inhibit wanted motions. Filippo Vella at ell solved the problem by maintaining the value of the AGMV constant when panning is detected.