Displaying the large number of bands in a hyper- spectral image (HSI) on a trichromatic monitor is important for HSI processing and analysis system. The visualized image shall convey as much information as possible from the original HSI and meanwhile facilitate image interpretation. However, most existing methods display HSIs in false color, which contradicts with user experience and expectation. In this paper, we propose a visualization approach based on constrained manifold learning, whose goal is to learn a visualized image that not only preserves the manifold structure of the HSI but also has natural colors. Manifold learning preserves the image structure by forcing pixels with similar signatures to be displayed with similar colors. A composite kernel is applied in manifold learning to incorporate both the spatial and spectral information of HSI in the embedded space. The colors of the output image are constrained by a corresponding natural-looking RGB image, which can either be generated from the HSI itself (e.g., band selection from the visible wavelength) or be captured by a separate device. Our method can be done at instance-level and feature-level. Instance-level learning directly obtains the RGB coordinates for the pixels in the HSI while feature-level learning learns an explicit mapping function from the high dimensional spectral space to the RGB space. Experimental results demonstrate the advantage of the proposed method in information preservation and natural color visualization.