RI UFLA (Universidade Federal de Lavras) >
Revistas UFLA >
Please use this identifier to cite or link to this item:
|Title: ||Efficient image fusion algorithm for multi-focus images using energy estimation|
|???metadata.dc.creator???: ||Zhang, Yingjie|
|Keywords: ||Image fusion|
Fusão de imagens
Função de energia
|Publisher: ||Editora da UFLA|
|Citation: ||ZHANG, Y .; GE, L. Efficient image fusion algorithm for multi-focus images using energy estimation. INFOCOMP: Journal of Computer Science, Lavras, v. 6, n. 1, p. 1-8, Mar. 2007.|
|Abstract: ||This paper presents a new image fusion approach for multi-focus image fusion. Considering that for the same scene, the energies of images captured on the focus and without the focuses are different, thus, fusion decision at some regions can be made based on the magnitudes of the energy. Certainly, fusion would be easily carried out once the source images have been segmented as a sequence of regions. However, the precise segmentation is usually difficult, and it is computationally expensive. To overcome the difficulties, a course mesh or cells, in this scheme, is defined in the image domain, each element corresponding to a set of pixels of the image on a rectangle window, which is approximated as regions. After that, fusion decision on the regions may be made by measuring the magnitudes of energy. Furthermore, note that the size of the regions may affect the fusion performance, particularly at the region boundaries, but it can be reduced by selecting finer mesh, and the cells may be spitted furthermore by using the marching squares algorithm to improve the fusion quality. By numeric experiments, it has been demonstrated that the proposed method could achieve similar fusion results compared with that obtained by the multi-scale decomposition-based image fusion algorithm, and with low computational cost.|
|Other Identifiers: ||http://www.dcc.ufla.br/infocomp/index.php/INFOCOMP/article/view/363|
|Appears in Collections:||Infocomp|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.