A modified multiscale fusion method combining context based on weight values for image segmentation is introduced. After the implement of the raw multiscale segmentation based on wavelet-domain hidden Markov tree model, weights for individual dyadic square of the image under consideration at each different scale are set. Finally, raw multiscale segmentation results are fused with the context based on the corresponding weight values. Experimental results show that the method can gain a better performance than the existing technique.