A region based random multi-pixel search algorithm to improve the binary hologram reconstruction quality

• We proposed a new iterative method for computing binary CGHs to improve the quality of reconstructed image. Considering the correlation between pixels of the hologram, several pixels in the same region are randomly selected from binary CGHs. By inverting multiple related pixels simultaneously in each iteration, this method overcomes the problem that DBS algorithm easily fall into local optimum. And the correlations between pixels can be controlled through selecting the pixels from the same region. Compared with DBS, the mean square error (MSE) of reconstructed image in simulation has declined from 844.85 to 445.74, the peak signal to noise ratio (PSNR) has increased from 18.86 to 21.64 and structural similarity (SSIM) has increased from 0.336 to 0.459. The reconstructed image in experiment presents fewer noise points and clearer outline. The numerical and experimental results show that this method can obtain higher reconstruction quality and lower distortion binary CGHs compared with other iterative methods. Direct binary search (DBS) is an effective method for the generation of binary computer-generated holograms (CGHs), which has been widely used in wavefront synthesis, optical tweezers and 3D display. However, the correlation of pixels in holograms is neglected in DBS, resulting the algorithm easily converging to local minimum and leading to an unsatisfactory reconstruction quality. To ameliorate this disadvantage, we proposed a new algorithm by randomly selecting multiple related pixels in the same region to generate hologram. Compared with DBS, the mean square error (MSE) of reconstructed image in simulation has declined from 844.85 to 445.74, the peak signal to noise ratio (PSNR) has increased from 18.86 to 21.64 and structural similarity (SSIM) has increased from 0.336 to 0.459. The reconstructed image in experiment shows fewer noise points and clearer outline. Both the numerical and experimental results demonstrate that the reconstruction quality can be improved by this algorithm, which will exhibit great applications in the fields of data storage, microscopy, and true sense dynamic 3D display.