Research On Super-resolution Reconstruction Algorithm Of Deep Undercooling Melt Image Under Microgravity Condition

Without crystalline nucleus and without gravity or microgravity,the liquid metal will reach a deep undercooling state,and the melt in this state will exhibit its characteristic properties.In order to simulate the gravity free or microgravity environment on the ground and study the properties of deep undercooling melts,researchers have set up an experimental environment for electrostatic suspension and vacuum drop tubes,and used high-speed cameras to capture images of deep undercooling melts during the fall.Due to the limitation of the high-speed camera itself,the resolution of the captured image is very low,which has an impact on the study of the properties of the deep undercooling melt.Therefore,digital image processing technology is needed to reconstruct the image,enrich the image detail and increase the image resolution.Nowadays,the existing image super-resolution(SR)method cannot restore the high-resolution(HR)image itself better from corresponding low-resolution(LR)image,and is not suitable for reconstructing the texture details of the deep undercooling melt.To solve this problem,we propose a method based on adaptive mixed sample and low-rank matrix decomposition optimization(AMS-LMDO)for single image SR.Complementary prior knowledge provided by the external sample library and the internal sample library was made full use of to solve this extremely ill-posed SR problem.In the proposed method,the prior knowledge provided by the mixed sample library is more suitable for deep undercooling melt image SR than the prior knowledge provided by any individual sample library.After reconstructing the image,low-rank matrix decomposition model was used to optimize HR image with erroneous information,removing sparse and uncorrelated errors,and making the resulting HR image closer to the original image.The simulation results show that compared with the current popular methods,the proposed method not only has a good effect in the reconstruction of general images,but also can restore inherent high frequency details of the undercooling melt.Compared with the well-recognized YangSR algorithm,the PSNR index of the method for undercooling melt image reconstruction is improved by about 1.86dB,and the SSIM index is improved by about 0.06.Compared with other advanced image super-resolution reconstruction methods,the reconstruction results of this method have different degrees of improvement in PSNR and SSIM.It follows that for the undercooling melt image discussed in this paper,the method can more clearly reconstruct the special high frequency details of the image.