Research On Methods Of Enhancement For Reconstructed Image Of Digital Holography

In recent years, accompany with improvement on the speed of personal computer andhigh-perfomanced CCD coming out, the overall research on digital holography is beginning.Thistechnology is applied to the domains of biomedical science,3-D object recognition, smalldisplacement measurement, vibration analysis, surface interference detection. The reconstructedimage of digital holography is the key in the real application, which has an effect on the accurancy ofmeasurement. The image suffering from zero-order image and speckle noise is vague. Aiming at theabove-mentioned problems, this article discusses how to improve the quality of reconstructed imageof digital holography, proposes a series of methods on noise suppression of digital holography image,eliminating zero-order image, noise suppression and enhancement of reconstructed image of digitalholography. The main tasks are below:Firstly, a noise suppression method of digital holography on modeling NSCT coefficients isintroduced. The method transforms the digital holography via NSCT. Under different scales，coefficients are fitted by different distribution models. Some factors are estimated by expectationmaximization (EM) and Bayesian estimation. Compared with classic methods such as Lee filtering,method of wavelet domain and Gaussian mixture model and method of contourlet domain andGaussian mixture model, the method proposed in this paper has advantages on visual effect and edgedetails.Secondly, a new method for eliminating zero-order image in digital holography based oncontourlet transform is presented. Digital holography is decomposed by contourlet transform, andthen the low frequency is removed. The new digital holography image is obtained by inversecontourlet transform. Experimental results show that, compared with the method of spatial filtering,the method of frequency domain filtering, the method of Laplacian filtering and the method ofeliminating zero-order image based on wavelet, the proposed method can eliminate zero-order imagebetter, what’s more, the real image can also be strengthened in some way.Then，a noise suppression method of reconstructed image of digital holography based oncomplex contourlet transform and anisotropic diffusion is proposed. Firstly, reconstructed image ofdigital holography is decomposed by complex contourlet transform. Then, according to thecharacteristics of coefficients, the appropriate anisotropic diffusion models are used. Finally, theimage is synthesized. Experimental results show that, compared with the method of wavelet thresholdshrinkage combined with total variation model, and the method of Laplacian pyramid-based nonlinear diffusion, the proposed method has better noise suppression effect while preserving details ofreconstructed image of digital holography.Next, a new enhancement algorithm for reconstructed image of digital holography based ondual-tree complex wavelet and fuzzy-logic is proposed. Firstly, digital holography image isdecomposed by dual-tree complex wavelet, getting the coefficients of low frequency and highfrequency. Then the coefficients of low frequency and high frequency are processed by usinggeneralized fuzzy operator. In mapping to ordinary fuzzy set, some parameters are self adapted byedge division theory. Then the inverse transform of dual-tree complex wavelet is used in coefficientmatrixs to enhance the image. The experimental results show this method has higher efficiency andbetter visual effect.Finally，an adaptive enhancement method of reconstructed image of digital holography incontourlet domain based on chaotic particle swarm optimization (CPSO) is proposed. Firstly,reconstructed image of digital holography is decomposed by contourlet transform, and the contrast isenhanced by linear gray stretch. Then the coefficients of noisy bandpass directional subbands areadjusted by nonlinear gain function. Taking three indexes such as contrast, definition andsignal-to-noise ratio into account, the integrated quantitative evaluation function is used as the fitnessfunction of chaotic particle swarm optimization algorithm. And the optimal parameters involved in theenhancement method in spatial domain and the nonlinear gain function for adjustment of bandpassdirectional subbands coefficients based on contourlet, are found by chaotic particle swarmoptimization algorithm. A large number of experimental results of reconstructed image of digitalholography show that, compared with the method of two-dimensional histogram equalization, themethod of second-generation wavelet, the method of stationary wavelet transform, the method ofcurvelet transform, the proposed method improves the contrast of the enhanced reconstructed imageof digital holography, increases the definition, reduces the noise and has a better overall visual effect.