De-interlacing Algorithm Based On Context

With a rapid development of new display devices and television format, there are lots of different standard formats in video field. It is essential to convert different video formats in order to achieve communication. In today’s TV broadcasting system, most of the video source is still interlaced scan. Interlaced scan can reduce the bandwidth, but also cause crawling, the screen flashes, blurred and jagged. De-interlacing, which is a key technology for vedio format conversion technology, is the basis for other format conversion technology. The so-called’de-interlacing’is the conversion technology from interlaced to progressive scan. In this paper, aiming at the changes on edges of the video sequence information, we focuses on the intra-field de-interlacing method based on context and the adaptive algorithm based on the motion detection. The main innovative reaserch work mainly includes the following aspects:(1) This paper proposed an adaptively single-field de-interlacing method based on edge-direction. This algorithm uses the geometric duality between low-resolution interlaced images and high-resolution progressive images to estimate the de-interlacing images. In order to improve the accuracy of the predicted model, we adapt the correlation coefficient for selecting training samples and predictive model. The experimental results show that this algorithm obtains high PSNR and good visual effects compared with traditional methods, especially in the edge.(2) This paper proposed an auto-regressive single-field de-interlacing method based on context. This algorithm is mainly to improve shortcoming that hign-resolution predictive model is estimated in one-way in edge-direction de-interlacing method. Improvement idea is to estimate low-resolution predictive model in two-way with estimated high-resolution model. By adapting edge detector, the tranditional intra-field de-interlacing method is conbined with our auto-regressive de-interlacing method to construct a new adaptive de-interlacing algorithm. The experimental results show that this algorithm can ensure a high PSNR value while reducing the computational load in auto-regressive de-interlacing method.(3) This paper proposed a new de-interlacing method based on motion compensation which combined intra-field and inter-field de-interlacing method. This algorithm first uses the motion detection to classify image for different motion states, then adopts different methods to de-interlace. In this paper, motion states are classified into three types, such as static, slow motion and fast motion. For static state, we adopt the average of pixels in corresponding positons around the before and post field. For slow motion state, we adopt a hybrid de-interlacing method which motion-compensated de-interlacing method is combined with the traditonal intra-field method. And for fast motion state, due to the inaccuracy of motion estimation, we use traditional single-field de-interlacing method. The experimental results show that this algorithm can obtain better subjective and objective results.