An Enhanced Algorithm for Automatically Capturing the Desired Moment in Photography

Digital photography has become ubiquitous in daily life, driven by the widespread use of smartphones. However, capturing the desired moment remains a challenge, particularly in scenarios involving rapid or fleeting motion. To address this, the previous approach introduced a method that enables cameras to autonomously plan and capture optimal moments with minimal user interaction [3]. The approach utilizes voice or text input to describe the user's intended photographic outcome, continuously evaluating the similarity between the provided description and the live preview image in a shared vector space. The preview image with the highest similarity score is automatically selected and captured as the optimal shot. Despite the effectiveness of this method, accuracy may be affected by various distracting factors in the scene. In this paper, we propose an enhanced algorithm that significantly improves accuracy by incorporating a dynamic sliding window to compare a continuous sequence of frames against the textual description, rather than individual frames. Experimental results demonstrate that the enhanced approach provides more robust accuracy, particularly in challenging scenes with various distractions.