The most significant problem may be undesirable effects for the spectral signatures of fused images as well as the benefits of using fused images mostly compared to their source images were acquired at the same time by one sensor. They may or may not be suitable for the fusion of other images. It becomes therefore increasingly important to investigate techniques that allow multi-sensor, multi-date image fusion to make final conclusions can be drawn on the most suitable method of fusion. So, In this study we present a new method Segmentation Fusion method (SF) for remotely sensed images is presented by considering the physical characteristics of sensors, which uses a feature level processing paradigm. In a particularly, attempts to test the proposed method performance on 10 multi-sensor images and comparing it with different fusion techniques for estimating the quality and degree of information improvement quantitatively by using various spatial and spectral metrics.