io_output · Graph boundary output
Graph boundary for embedded execution: the host collects the graph result from io_output after the run.
Graph boundary for embedded execution: the host collects the graph result from io_output after the run.
Edge-preserving denoise: weights by spatial distance and color difference together.
Box-style mean blur: averages the kernel window — the fastest way to smooth.
Box filter: sums the kernel window (optionally normalized) — the generic form of the mean blur.
Morphological dilation: max over the window — bright regions grow, gaps close.
Morphological erosion: min over the window — bright regions shrink, white specks vanish.
2D convolution with any kernel you write down as a numeric matrix.
Directional texture filtering with Gabor kernels — a classic for fingerprint/fabric analysis.
Gaussian-weighted blur: natural smoothing, the default choice for denoising.
Reads one image from disk (png / jpg / bmp / tiff / webp …) as the graph’s input source.