Image filters work through a mathematical process called convolution, where a small matrix called a kernel slides across every pixel of the image, computing a weighted sum of each pixel and its neighbors to produce the output value. The values in the kernel determine what visual effect the filter produces, and understanding this simple mechanism reveals how diverse effects like blur, sharpen, and edge detection all emerge from the same fundamental operation.

Gaussian blur, the most widely used blur algorithm, derives its kernel values from the Gaussian (normal) distribution, the familiar bell curve from statistics. The two-dimensional Gaussian function assigns the highest weight to the center pixel and progressively lower weights to surrounding pixels based on their distance, with the falloff rate controlled by the sigma parameter. A larger sigma means a wider bell curve and therefore a stronger blur effect. The Gaussian distribution is mathematically elegant because blurring an image with a Gaussian and then blurring the result with another Gaussian is equivalent to a single blur with a combined sigma, and a 2D Gaussian blur can be decomposed into two sequential 1D passes (horizontal then vertical) for dramatically faster computation.

Sharpening works by enhancing edges and fine detail, essentially the opposite of blurring. The most common approach is unsharp masking, which despite its counterintuitive name actually increases sharpness. The algorithm creates a blurred copy of the image, subtracts it from the original to isolate the high-frequency detail, then adds that detail back to the original at an amplified intensity. The sharpening kernel in convolution form typically has a positive center value greater than 1 surrounded by negative values, which mathematically amplifies differences between each pixel and its neighbors.

Color space transformations underpin many filter effects. Converting from RGB to HSL (Hue, Saturation, Lightness) enables intuitive adjustments: rotating the hue shifts all colors around the color wheel, reducing saturation moves toward grayscale, and adjusting lightness brightens or darkens. The sepia effect is achieved by converting to grayscale (using perceptual weights of approximately 0.299 for red, 0.587 for green, and 0.114 for blue to match human brightness perception) and then mapping the grayscale values through a warm brown color ramp. Vintage effects combine multiple transformations: slight desaturation, warm color temperature shift, reduced contrast, and sometimes vignetting (darkening corners) to simulate the look of aged photographic prints.