Abstract: With the upgrading of hardware equipment such as cameras and the improvement of audience's aesthetic level, green screen keying technology under high resolution has become very important. In order to achieve accurate and efficient green screen keying in high resolution images, a high resolution automatic real-time green screen keying method based on attention mechanism is proposed. The proposed method is divided into two stages. In the first stage, only the RGB green screen image is used as input, and the attention mechanism is used to accurately extract the foreground object, without manual participation or the generation of approximate background image and other pre-processing operations, which realizes automation. In the second stage, the depth guided filter module is added for high resolution processing, and the real-time processing efficiency under high resolution is realized. The experimental results on the green screen data set show that the proposed method can achieve a processing speed of about 80.2 frames /s under 2K resolution video under the premise of ensuring good green screen keying accuracy.