Practical image fusion method based on spectral mixture analysis

Abstract

Conventional image fusion algorithm, such as IHS, SVR, PCS, etc., may show some defects in inheriting the higher-spectral information embedded in the original lower-spatial resolution MS image. A fusion method based on spectral mixture analysis (FSMA) was proposed in previous study, which has potential in solving this problem. While published results are limited to well-behaved simulated data where the endmembers are known a priori and the FSMA method will not work well when applying to real remotely sensed images because the estimated reflectance ranging in panchromatic band derived from MS bands cannot be treated as the real panchromatic values. In this paper, an improved image fusion method based on spectral mixture analysis (IFSMA) is proposed, in which the original FSMA method was extended to real remotely sensed images by modifying the objective function of the constrained nonlinear optimization expressions. It was compared with the original FSMA, Zhang’s SVR, PCS and IHS method, and results indicated that the IFSMA method was superior to other methods in preserving the spectral and spatial information.