VLSI architectures for real-time moving image resampling

This paper presents a solution for real-time image resampling, using dedicated VLSI architectures. The spatial image resampling algorithm is based on the use of two levels for the interpolation and relies on a tradeoff between storage and calculation of the filter coefficients, improving the performance-cost tradeoff of the hardware. Efficient motion compensated temporal resampling is carried out with a high performance motion estimation algorithm. Multiprediction, multiresolution and hierarchical approaches lead to an overall reduced computation power compared to classical block-matching algorithms