Something Weird Happened When We Downsize an Image

When naively downsampling an image by removing every other row and column, strange visual artifacts appear. This is aliasing — a phenomenon explained through signal processing theory. The Nyquist sampling theorem states you must sample at least twice the highest frequency in a signal to avoid aliasing. For images, the fix is anti-aliasing: apply a low-pass filter (like a Gaussian blur) before downsampling to remove high-frequency content. The post also covers reconstruction filters (sinc function, Lanczos resampling, bicubic interpolation) and Gaussian pyramids, which are used in scale-space analysis, feature extraction, and object detection.