Research On Algorithm Of Wavelet Neural Network For Computer-Generated Hologram Compression

Holographic display technique is considered to be one of the most ideal and popular 3D display technologies for displaying the images of objects with all viewpoints and various distances realistically so as to meet the needs of human visual perception. However, very large data quantity contained in hologram makes it difficult to display in real time and complete efficient transmission, analysis and storage so that it has seriously hindered application development of the three-dimensional dynamic holographic display technology. Therefore, it is urgent for hologram to find a faster compression method possessing higher degree compression and better image fidelity.Based on the characteristics of the fringe information of hologram and existing hologram compression algorithms, an algorithm of Computer-generated Hologram (CGH) compression is presented by use of wavelet neural network (WNN). By constructing the mapping relationship of input and output data, huge holographic data is stored in the network in the form of a small amount of connection weights and other parameters. This algorithm not only can compress holographic data greatly, but also obtain high quality reconstruction image. In the experiment, the model structure of hologram compression algorithm is designed on the basis of wavelet neural network and the simulation program is written to verify with valuable data results.In view of the slow convergence speed, easy falling into local minima, poor stability and difficulty on the choice of initial value in the training of wavelet neural network, the Genetic Algorithm (GA) is introduced to optimize the weights parameters, considering that genetic algorithm is a random search algorithm with reference species evolution mechanism in nature, and it has the advantages of parallel processing and global optimization ability. And the hologram compression steps are also constructed in the model using wavelet neural network (WNN) after genetic algorithm (GA) optimization. The experimental results show that the compression algorithm not only can get faster speed of convergence and better ability for global searching, but also perform more outstanding robustness and more excellent stability. In order to verify the compression method is reasonable and efficient, comparative experiments have been done on groups of images under the same experimental environment with several current compression algorithms such as DCT, DWT and BP algorithm. The experiments demonstrated that when the compression rate is as low as 1.56%, Peak Signal to Noise Ratio (PSNR) of the modified algorithm is the best among other existing algorithms, and Mean Squared Error (MSE) is lowest. Moreover, the details of reproduced image could be seen more clearly and the proposed method preserves better robustness property. It fully shows that the proposed algorithm and its modified algorithm are suitable for hologram compression with the nonlinear, unstable, dynamic information inside.