Efficient memory access methods for framebuffer-less video processing applications

Mask processing is commonly used in many video/image processing functions such as scaling, noise reduction, and edge enhancement by referring several pixels in the adjacent area to generate a processed pixel. Embedded memory therefore occupies large area and consumes power to store and process pixel data, especially for those low-cost applications without a framebuffer. In this paper, we design a scaling engine with the proposed pixel-based memory access method, which achieves 52% power reduction and 37% line buffer reductions as compared with the conventional line-based memory access method. In addition, it also reduces 10% power consumption for a general noise reduction processing because the pixel-based memory access method can efficiently reduce the memory activity.