DMA Image Processing Eliminates Frame Buffer

The patent segments the image processing into distinct stages: fetching image tiles via DMA channels, performing color space conversion on individual tiles, executing alpha blending operations, and managing multiple image streams independently. This segmentation allows each component to handle specific format conversions without requiring the entire system to support all format combinations simultaneously, thereby improving adaptability while managing complexity.

The patent introduces intermediate buffer memory and control logic as mediators between the DMA fetching module and the alpha blending unit. These intermediaries temporarily store image tiles in standardized formats, allowing the system to accept images in various formats (YUV42x Planar, RGB, etc.) and convert them to a common internal representation before blending, thus enhancing format compatibility without directly coupling all format handlers.

The patent employs dynamic memory allocation where buffer memory is allocated only when and where needed during the alpha blending process. The system dynamically manages the size and location of intermediate buffers based on the specific image formats and resolutions being processed, rather than allocating fixed large buffers. This dynamic approach ensures sufficient memory for reliable processing while minimizing unused memory allocations and associated power consumption.

The patent extracts the buffering function from a dedicated external storage frame buffer and integrates it into the alpha blending pipeline using shared buffer memory. By taking out the requirement for a separate large frame buffer and using smaller, strategically placed intermediate buffers within the processing pipeline, the system maintains processing reliability while reducing overall memory usage and power consumption.

The patent merges the color space conversion, scaling, and alpha blending operations into a unified processing pipeline that operates on image tiles in a single pass. Instead of using a temporary frame buffer to hold entire intermediate images for later processing, the system combines these operations so that tiles are converted, scaled, blended, and output in one continuous flow. This merging eliminates the time penalty associated with writing to and reading from temporary buffers.

The patent changes the processing dimension from whole-image operations requiring temporary storage to tile-level stream processing. By processing images as a continuous stream of tiles rather than loading entire images into temporary buffers, the system achieves frame bufferless operation. The tile-based approach in another dimension (spatial segmentation) allows progressive processing without temporal delays from buffer I/O operations.