Research On Multi-focus Image Fusion Algorithms

Multi-focus image fusion is to fuse a completely clear image with a set of images of the same scene and under the same imaging conditions which are with different focus points. The efficiency of the image information and the reliability of the system detecting and recognizing the objects can be improved through multi-focus image fusion. This dissertation studies the multi-focus image fusion algorithms, and the most important research and findings are as follows:(1) The typical image fusion methods in airspace are analyzed and proved, and an improved gradient operator algorithm is proposed in this dissertation. This algorithm can fuse multi-focus images on the basis of confirm the real gradient direction of the images. The results of the simulated experiments indicate the algorithm has better performance than traditional fusion algorithms in airspace.(2) Analyze the elements of the Fourier transform and the wavelet transform, indicates the advantages of wavelet transform which is compared with Fourier transform in the field of image fusion. Study the multi-focus image fusion based on wavelet transform, and some typical multi-focus image fusion algorithms are discussed separately from the high frequency and low frequency view. Analyze the fusing effect of choosing different wavelet bases and decomposed levels. Propose an improved wavelet algorithm using edge detection which is based on the character of wavelet function and the characteristic of multi-focus images. The results of subjective observation and objective experiment data show that the improved wavelet image fusion algorithm has a good performance in fusing multi-focus images.(3) Study the evaluation parameters of the image fusion, and conclude the performance indicators which can be applied to evaluate the multi-focus image fusion. This dissertation adopts some parameters such as entropy, mean gray level, root-mean-square deviation, resolution, and spatial frequency and so on as the evaluation parameters. These parameters evaluate the airspace and wavelet transform image fusion algorithms mentioned in this dissertation.