High Resolution Reconstruction for Dual-Camera System

To obtain the wide FOV target image with high resolution, a high resolution reconstruction for dual-camera system was proposed. Firstly, the high frequency feature was learnt based on high/low resolution image pair with narrow FOV, and a linear regression function was trained to represent the mapping of high frequency feature between high resolution and low resolution. Then, the super resolution reconstruction of wide FOV image was realized in wavelet domain. At last, ten groups of simulation experiments were carried out to verify proposed high resolution for dual-camera system, comparing with popular bicubic interpolation, wavelet based interpolation and self-similarities based reconstruction. Experimental results showed that the proposed approach performs best with small time cost, whose PSNR was higher than bibubic interpolation about 0.26.