Computer-Generated Hologram Encoding Using Precomputed Lookup Tables

The patent applies preliminary action by pre-calculating and storing encoding tables for different light frequencies before actual CGH generation. The system pre-processes the encoding requirements for various frequencies and stores them in lookup tables, so that during runtime, the encoding can be performed by simple table retrieval rather than complex real-time calculation. This resolves the contradiction by preparing the encoding data in advance, improving image clarity while keeping the actual generation process simple.

The patent introduces an intermediary encoding table that acts as a mediator between the input image data and the final CGH output. Instead of directly encoding complex multi-frequency light data, the system uses pre-computed encoding tables that translate image data into appropriate holographic patterns for each frequency. This intermediary structure simplifies the encoding process while ensuring all frequency components are properly directed to reach the observer's eye lens.

The patent applies preliminary action by pre-training the deep learning network offline and storing its learned encoding patterns in lookup tables. The complex computational work of learning optimal encoding strategies for different frequencies is performed in advance during network training, not during actual CGH generation. During runtime, the system only needs to retrieve pre-computed encoding values from the tables, dramatically reducing computational time while maintaining high encoding precision.

The patent uses copying by creating simplified representations of the complex deep learning encoding process in the form of lookup tables. Instead of executing the full deep learning network during CGH generation, the system copies the essential encoding patterns learned by the network into tabular form. This allows the system to replicate the high precision of deep learning encoding while avoiding its computational overhead during actual operation.

The patent applies preliminary action by pre-determining the optimal number of iterations and convergence criteria during system setup. The iterative refinement process is configured with predetermined parameters that prevent excessive iterations, allowing the system to achieve sufficient encoding accuracy without unnecessary processing time. The refinement process stops once a pre-established accuracy threshold is met, balancing precision and duration.

The patent applies partial action by performing iterative refinement only when necessary - specifically, only for frequency components that require correction to ensure they reach the observer's eye lens. Not all encoding operations require full iterative refinement; the system selectively applies refinement to specific frequency components or data portions that need it, reducing overall processing duration while maintaining necessary encoding accuracy.