Signal Reconstruction Techniques Applied to High Frame Rate Sequences

Nowadays there are specific video equipments which allow acquisition at very high frame rates, e.g., up to 5000 fps. When observing a high frame rate sequence, it becomes noticeable that a great percentage of the frames' pixels values do not suffer great changes along the sequence. Hence, if a pixel's value remains the same, then a unique sample is enough to represent that particular pixel over several frames. Only pixels that suffer great variations require a higher number of samples to be correctly represented. This work reports a methodology that combines compression and reconstruction techniques to evaluate an alternative to the traditional compression methods 25/30 fps. In the proposed method, sets of pixels are extracted along the temporal direction and by exploiting the high temporal correlation of the signals, a proper decimation process is applied in order to create a sequence with a lower number of frames and consequently smaller file size. The main objective is to propose a new compression methodology that will take advantage of the signal characteristics using both compression and reconstruction techniques.