HDR Video Editing with Mask-Guided Alpha Map Allocation

The patent segments the blending process into two distinct stages: a decision stage that determines whether blending is needed (using minimal mask information), and an execution stage that performs actual blending (using full alpha information). This segmentation allows the system to maintain high blending accuracy when needed while avoiding unnecessary memory allocation when blending is not required, directly resolving the contradiction between blending precision and memory consumption.

The patent implements partial action by storing only mask information (partial data) during the decision stage instead of complete alpha information. This partial storage approach consumes minimal memory while still enabling accurate blending decisions. The full alpha information is allocated temporarily only when blending is actually executed, ensuring that memory is used efficiently without compromising blending accuracy when required.

The patent implements dynamic memory allocation where the system adaptively adjusts memory usage based on the actual editing requirements. During the decision stage, minimal memory is used for mask operations. When blending is determined to be necessary, the system dynamically allocates alpha maps only for those specific frames or regions. This dynamic approach maintains high editing performance while avoiding the complexity of permanently allocating large memory buffers, as memory is allocated on-demand rather than statically.

The system performs self-service by automatically determining when blending is needed through mask information analysis, and autonomously allocating or deallocating alpha maps without requiring external intervention or complex memory management protocols. This self-managing approach simplifies the overall system architecture while maintaining high editing performance, as the system serves its own memory management needs efficiently.

The patent applies preliminary action by pre-storing mask information that encodes blending decisions before actual blending operations. These masks are generated during video decoding or preprocessing stages, preparing the data structure for efficient later use. This preliminary preparation maintains memory efficiency during storage while enabling flexible blending operations when needed, as the pre-computed masks can be quickly referenced and expanded into full alpha maps only when required, thus preserving both memory efficiency and blending flexibility.

The patent uses mask information as an intermediary between the compact decision-making stage and the full-featured blending execution stage. The mask serves as a lightweight mediator that contains essential blending guidance without requiring full alpha map data. This intermediary structure maintains memory efficiency during the decision phase while enabling complete blending flexibility during execution, as the mask can be expanded or interpreted into various blending modes without losing the underlying flexibility options.