Video Blur Effect Using Sequential Box Filters

The patent segments the blur calculation process into multiple sequential box filter passes. Instead of performing a single complex Gaussian convolution, the method applies multiple simpler box filters in sequence, where each pass processes the output of the previous pass. This segmentation reduces the computational complexity from O(n²) multiplications to O(n) additions per pass, while still achieving Gaussian-like blur quality through the cumulative effect of multiple passes.

The patent uses simple box filters with unity weights as intermediate processing stages rather than complex Gaussian kernels. Each box filter is a computationally inexpensive operation that can be quickly applied and discarded, with the final Gaussian-like effect emerging from the composition of multiple such simple operations rather than from a single complex filter.

The patent divides the blur operation into multiple sequential passes of simple box filters. Each pass processes the video data quickly using only addition operations, and the cumulative effect of multiple passes achieves the desired Gaussian blur quality. This segmentation enables real-time processing by maintaining low computational complexity per frame while still delivering high visual quality.

The patent applies box filters sequentially in multiple continuous passes through the video data. Each pass builds upon the previous pass's output, creating a continuous processing pipeline that efficiently transforms the video. This continuous multi-pass approach maintains processing speed by avoiding idle time between operations while accumulating the blur effect progressively.