| dc.description.abstract | High- definition display has become the mainstream of media player, but most of the current distribution of video content is not premise of making High-definition video quality, using scaler to enlarge the input video resolution is a common solution today, the core of scaler shall be interpolator.
This paper developed a probabilistic neural network-based intelligent video interpolator. This interpolator analysis of image in the region around the each interpolation point sharpness, according to the evaluation value of sharpness as a single neuron input to adjust the smoothing parameter of PNN interpolator, finally using PNN interpolator to do it. The probabilistic neural network-based intelligent video interpolator has to adapt to different characteristics of video interpolation region, it provides more sharp effect near the edge, and in the smooth region, to enhance smooth effect is provide. Compared with conventional video scaler, the system shows a very excellent scaled image quality.
This intelligent video scaler has verified its functional correctness and performance in the FPGA test platform. Apart from implement PNN interpolator hardware circuit, we also completed the memory controller, LCD controller, image processor, finally design a pipeline controller for each of the parallel hardware architecture module integrated into the video scaling chip. The chip uses only 2147 Logic Elements and 241k bits of memory on FPGA hardware resources, and can reach 30 frames 1920x1080 images per second interpolation rate, indicating the system has satisfactory performance. The research results will be used in digital television, high-definition video surveillance and portable media player devices.
| en_US |