Iterative second-order derivative-based deinterlacing algorithm

In this paper, we introduce an iterative second-order derivative-based intra-field deinterlacing algorithm (ISDD). The proposed ISDD is composed of two steps: (1) preprocessing by a fast second-order derivative-based deinterlacing algorithm (FSDD), (2) iteratively refining the interpolated pixel by minimizing a mass second-order function. The function includes three components: the derivative difference, directional derivative, and isophote smoothing. The derivative difference is used to detect any intensity discontinuity in an image; the directional derivative is a supplement to the derivative difference to avoid oversmoothing; and isophote smoothing can reduce artifacts. The proposed ISDD can preserve image features and reduce the artifacts caused by conventional deinterlacing methods. Compared with existing deinterlacing algorithms, the proposed algorithm improves the peak signal-to-noise ratio while maintaining high efficiency.

► We study iterative second-order derivative-based intra-field deinterlacing method.
► ISDD is adaptive to the image feature for selecting the direction using derivatives.
► ISDD iteratively corrects the pre-processed image using FSDD.
► ISDD provides good reconstruction performance and requires a low computational cost.