Video deinterlacing with dynamic motion analysis and scene cut detection

Abstract

In this paper we will propose a new deinterlacing method which uses a scene cut detection technique. The proposed method first performs scene cut detection on interlaced video sequences. If a scene cut is detected between the current field and the reference field, we suggest reducing the interpolation error caused by incorrect motion compensation. On the other hand, if no scene cut is detected, motion detection is performed on the current field. The motion detection based on block variance classifies a block of the current field into one of three groups, namely: static, semi static and non-static. Then, the modified bilateral motion estimation is performed for non-static blocks, simple bilinear interpolation is used for semi-static blocks and finally static blocks will inherit the original pixels from the reference field. In the experiments for the subjective evaluation, the proposed method has shown acceptable results in terms of peak signal-to-noise ratio and computation time consumption.