The traditional detection methods for these regions of interest in high resolution remote sensing image generally search the whole image on the basis of prior knowledge, which leads to high computational complexity. A new fast detection algorithm based on human visual characteristic for these regions of interest in high resolution remote sensing image is proposed. The new algorithm based on visual attention model applies spatial dimension reduction strategy to confirm these focuses of visual attention. According to the positions of focuses of visual attention, the new algorithm describes these relevant regions of interest in the original remote sensing image. The experimental results show that the new algorithm could not only have the lower computational complexity, but also avoid image segmentation and feature detection for the whole image and improve the detection efficiency of regions of interest in high resolution remote sensing image.