Different focuses image fusion with directional support value transform

To recover an everywhere-in-focus image, the multi-scale analysis image fusion is a classical method. Within these multi-resolution decompositions, the salient feature components image sequences with the largest magnitude are selected at each pixel location and finally, the fused image can be recovered from the decomposed components image sequences. Under the LS-SVM framework, salient features underlying image are represented by support values, and support value transform (SVT) has been developed for image fusion. To represent edges more efficiently, we analyze image under the weighted mapping LS-SVM framework, and deduce the directional support value filters and develop directional SVT to separate edges with different orientations in each image. The parameters of the weighted mapping LS-SVM for directional support value filter is optimized for the different focuses image fusion. Experimental results demonstrate that the proposed method can give superior results in the fused images comparing to the standard SVT and the discrete wavelet transform methods.