Fast Image Fusion In Spatial Domain

Image fusion aims at exploiting complementary information in multimodal images to create a single composite image with extended information content..Compared to transform domain fusion,the advantage of the spatial domain fusion method is that it directly processes pixels rather than processes transform domain coefficients,which can preserve the details of different scales of the image rather than the limited scales determined by the decomposition levels,and also avoid the additional computational burden of multi-scale decomposition.After extensive research on spatial domain fusion techniques,in this paper,we use fast filtering for image fusion in the spatial domain.First of all,because the gradient image contains rich texture and boundary information of image structure,we use the gradient magnitude as a measure of the saliency to detect the contrast and the regional sharpness of the image,and the use of forward difference can quickly realize a discrete gradient.Secondly,a morphological closing operation is used to refine the gradient map since the salient structure detection may produce gaps at a continuous curve and holes in a homogeneous region.However,the dilation operation is a time-consuming procedure due to calculating the maximum value in the sliding structuring element.To accelerate the procedure,we use the boxfilter to approximate convolution operation instead of the dilation operation.The intrinsic structures in multimodal source images cam be extracted by using the fast morphological filtering.Then,we propose a fast structure-preserving filter to process weighted map that contains different scales of intrinsic structures,and the weighted map is enhanced by smoothing the details and preserving the structure.The filter mainly uses the linear approximation principle,and the boxfilter is used to calculate the average of pixels within a certain window,because it uses the integral image principle,the structurepreserving filter has a low complexity.Finally,the fused image is obtained by the weight averaging on input images.To verify the superiority of the algorithm,in this paper,three types of input images,four quality evaluation metrics and five state-of-the-art comparing algorithms are used in experiment.Experimental results show that the final fusion image of the proposed algorithm achieves satisfactory results in visual effects.At the same time,the algorithm is superior to other image fusion algorithms in terms of objective evaluation and computational efficiency.