Multi-focus image fusion using local structure features with dynamic windows

Due to the limited depth-of-focus range, only the objects falling in the range of focus are sharp while others are blurred in images captured by general optical lens imaging devices. Usually image fusion techniques are employed to provide fully focused images. But, some unnatural flaws always appear in the object region or on the boundary in images fused by image fusion algorithms. Here, an algorithm is proposed to solve the problems mentioned above. While evaluating the focus measure of each pixel, firstly a local window size for a pixel is found depending on its location. Secondly, the pixel structure strength is obtained by calculating the gradient within the local window. Once the initial focus map is obtained, a morphological filter and the small region filter are used to make this map more smooth and spatial consistent. In the experiment, several commonly used multi-focus image sets are applied. The experiment results show that the proposed algorithm can reduce the unnatural artifacts and obtain the boundaries of objects more precisely.