For color image Super-resolution (SR) reconstruction, the common practice is to transform the images from the Trichromatic (RGB) color space to YUV or some other color spaces that separate the luminance from chrominance and saturation components, and implement SR reconstruction only on the luminance component (Y). As for the color components (U, V), usually directly use interpolation algorithms for enlargement. Although interpolation process is high-efficiency, it often produces “jaggy” and “ringing” artifacts. When the images are complex or the applications want better quality, this simple process cannot satisfactory. This paper proposes an approach using Projection Onto Convex Sets (POCS) theory, integrating SR reconstructed Y component into the local constraint to improve the edge performance of the reconstructed high-resolution U, V, and setting a global constraint to ensure the whole performance. Experimental results demonstrated that compared with the bicubic- interpolation method and the colorization-based U,V component process method which has the same focus as us, our algorithm can reconstruct the high-resolution U,V component with higher Peak Signal Noise Ratio (PSNR) and Structural Similarity (SSIM) and the efficiency of our approach is acceptable.