目的 由于光流估算的缺陷、噪声干扰以及现有运动注意力模型的局限性，导致运动注意力计算结果不能准确反映运动的显著性特征，制约了运动显著图的进一步应用。为提高运动注意力计算的准确性，提出一种基于时—空多尺度分析的运动注意力计算方法。方法 该方法根据视觉运动注意力来自于时—空运动反差的注意力形成机理构建运动注意力模型；通过时间尺度滤波去除噪声影响；鉴于视觉观测对尺度的依赖性，通过对视频帧的多尺度分解，在多个空间尺度进行运动注意力的计算，根据宏块像素值的相关系数大小对低尺度、中低尺度和原始尺度的运动注意力计算结果进行融合，得到最终的运动注意力显著图。结果 对多个视频测试序列的测试，测试结果表明，本文方法比同类方法更能真实有效地反映出视频场景中的运动显著性特征，大大提高了运动显著图的准确性。结论 为有效提高运动注意力计算不准确的问题，提出一种基于时—空多尺度分析的运动注意力计算方法，对于不同复杂视频运动场景，该方法能明显增强运动注意力计算的准确性，为视觉运动注意力的进一步应用奠定了良好基础。

Objective For the defect of optical estimation, noise and the limitations of existing motion attention model lead to the computation results of motion attention, which cannot accurately reflect the conspicuous characteristics of motion and constrain further application of motion conspicuous map. To improve the computation accuracy of motion attention, we suggest a target detection algorithm based on multi-scale motion attention analysis in this paper. Method According to the mechanism of visual attention, spatial-temporal motion attention model is built. Then noise influence is reduced by the time filtering. In view of the visual observation of scale dependence, the video frames are decomposed in multiple scales and the motion attention is also computed in spatial multiple scales in space. On the basis of the correlation coefficient of macro block pixel, low scale, middle scale, and the original scale, the motion attention computation results are fused to obtain the final motion attention map. Result The test result using different videos show the algorithm is more correct for motion attention than other algorithms and it greatly improves the accuracy of the motion attention map. Conclusion In order to improve the inaccurate computation of motion attention, we propose a motion attention computation algorithm based on spatio-temporal multi-scale analysis. For different complex motion video scenes, the proposed algorithm can obviously enhance the computation accuracy of motion attention and lay a good foundation for the further application of visual motion attention.