This paper presents a maximum a posteriori (MAP) based intra-field deinterlacing algorithm. In the proposed algorithm, we propose a hybrid approach composed by point-wise and patch-wise measurements. The estimation of the missing pixel is formulated as an MAP and minimizing the energy function. By utilizing Bayes theory and some prior knowledge, the missing pixel is estimated with a statistical-based approach and we model the residual of the images as Gaussian and Laplacian distribution. Under the MAP framework, the desired deinterlaced image corresponds to the optimal reconstruction given the interlaced low resolution image. Compared with existing deinterlacing algorithms, the proposed algorithm improves peak signal-to-noise-ratio and the structural similarity while maintaining high efficiency.
© 2014 IEEEPDF Article