By fusing a low spatial resolution hyperspectral image (LR-HSI) and a high spatial resolution RGB image (HR-RGB), hybrid-resolution hyperspectral imaging has been a popular framework for acquiring high spatial resolution hyperspectral images (HR-HSIs). Existing fusion methods always employ a known spectral response function (SRF) of the RGB camera to reconstruct the HR-HSI. The SRF is often limited or unavailable in practice, restricting the performance of existing methods. To address this problem, we propose a color space transfer-based fusion strategy that obtains HR-HSIs based on a hybrid resolution hyperspectral imaging system without measuring the SRF. Specifically, using a clustered-based backpropagation neural network, the HR-RGB is mapped into the CIE XYZ color space, and the HR-XYZ is obtained. In the CIE XYZ color space, its SRF is known; thus, the the SRF measurement is successfully skipped. To efficiently fuse the HR-XYZ and the LR-HSI, we propose a polynomial fusion model that estimates the ratio matrix between the target HR-HSI and the upsampled LR-HSI. Finally, the target HR-HSI is reconstructed by combining the ratio matrix and the unsampled LR-HSI. Experimental results on two public data sets and our real-world data sets show that the proposed method outperforms five state-of-the-art fusion methods.