Abstract: For the computational complexity of traditional non-sampling wavelet algorithm, a method of multi-focus image fusion based on fast non-sampling wavelet with the advantage of the characteristic of the processing on both space and frequency domain is proposed in this paper. Inverse transform is not needed as compared to previous non-sampling wavelet algorithms. Pixel value of fusion image is directly taken at the corresponding position of the source image based on the maximum principle of the sum of the absolute value of high frequency wavelet coefficients. Hence, there is a great improvement both on the real-time requirement of image processing and the fusion effect. Two comparisons, namely, comparison between the algorithm and six traditional non-sampling wavelet algorithms, comparison between fast non-sampling wavelet transform and traditional sampling wavelet transform are made in the work. The comparison results validate the advantage of the algorithm on effect and the time cost.

Key words: fast non-sampling, wavelet transform, fusion algorithm, space domain

摘要： 针对传统非抽样小波变换算法较复杂的缺点，结合空、频域处理上的特点，提出了一种基于快速非抽样小波变换的多聚焦图像融合算法。与之前基于非抽样小波变换的融合算法不同，该算法取消了反变换，它根据高频小波系数绝对值和取大原则，融合图像像素值直接在对应源图像的相应位置取值，从而大大提高了图像处理的实时性，改善了融合效果。通过与六种非抽样小波变换融合算法的比较，以及快速非抽样小波变换与非抽样小波变换的融合时间对比，直观地给出了该算法的效果和时间优势。

关键词: 快速非抽样, 小波变换, 融合算法, 空域