To accurately extract key frames from film and television videos, a key frame extraction algorithm is proposed that integrates K-means clustering and inter-frame similarity. Spatial difference measurement and perceptual hash measurement are integrated into video features to form pixel difference measurement. Threshold is set, pixel difference measurement are combined and histogram measurement to determine whether shot switching has occurred, and shot segmentation in film and television videos is achieved. The features of film and television videos are extracted, the initial cluster center position and number are dermined based on inter frame similarity and threshold, the initial cluster center is optimized using K-means, and the frames of the cluster center are extracted as key frames for film and television videos. The experimental results show that the proposed algorithm has significantly improved its fidelity and shot reconstruction ability, achieving accurate extraction of key frames in film and television videos, and can comprehensively describe the main content of the video.

K-means, interframe similarity, film and television videos, keyframe extraction, lens segmentation