Team
math

Convolution Calculator

Convolution is a fundamental operation in mathematics, signal processing, and probability theory that combines two sequences or functions to produce a third. Engineers use it for filtering signals, statisticians apply it to probability distributions, and computer vision relies on it for image processing. Enter up to 15 terms per sequence to compute the convolution and explore how this powerful tool shapes modern data analysis.

Last updated:

Creators Anna Szczepanek, PhD Mathematics
8,571 people find this calculator helpful

Understanding Convolution

Convolution is a mathematical operation that takes two input sequences and produces a single output sequence by systematically combining their elements. The operation is denoted using an asterisk symbol (∗), so the convolution of sequences a and b is written as a∗b.

At its core, convolution measures how one sequence modifies or filters another. This concept appears across diverse fields: in signal processing, convolution applies filters to audio or image data; in probability, it describes how independent random variables combine; in physics and engineering, it models how systems respond to inputs over time.

What makes convolution particularly useful is that it transforms a complex operation into a manageable computation—once you understand the mechanics, you can predict how signals will behave when filtered or how probability distributions will evolve.

The Convolution Formula

For two sequences a = {a₀, a₁, a₂, ...} and b = {b₀, b₁, b₂, ...}, the convolution produces a sequence c where each term is calculated by multiplying overlapping elements and summing the results. The index notation starts at zero, following standard programming convention.

cₙ = Σ(k=0 to n) aₖ × bₙ₋ₖ

where cₙ represents the n-th term of the result

  • cₙ — The n-th element of the output sequence
  • aₖ — The k-th element of the first input sequence
  • bₙ₋ₖ — The element of the second sequence at index (n − k)
  • n — The position in the output sequence

Step-by-Step Convolution Process

Computing convolution by hand follows a straightforward algorithm:

  • Step 1: Multiply the first element of sequence a by the first element of sequence b. This product becomes the first output term.
  • Step 2: For the next output term, compute pairs where the first sequence index increases and the second decreases: (a₀×b₁) + (a₁×b₀). Sum these products.
  • Step 3: Continue expanding: (a₀×b₂) + (a₁×b₁) + (a₂×b₀) for the third term, and so on.
  • Step 4: Repeat until you've generated as many output terms as needed. The output length equals the sum of input lengths minus one.

Example: Convolving [1, 2, 3] with [4, 5, 6]:
c₀ = 1×4 = 4
c₁ = (1×5) + (2×4) = 5 + 8 = 13
c₂ = (1×6) + (2×5) + (3×4) = 6 + 10 + 12 = 28
c₃ = (2×6) + (3×5) = 12 + 15 = 27
c₄ = 3×6 = 18
Result: [4, 13, 28, 27, 18]

Convolution in Continuous Functions

Beyond discrete sequences, convolution extends to continuous functions. For two real-valued functions f and g, their convolution is defined as an integral that blends the two functions across all possible offsets:

(f ∗ g)(x) = ∫₍₋∞₎^∞ f(t) × g(x − t) dt

This integral formulation is essential in signal processing, where f represents an input signal and g acts as a filter. The integral computes a weighted average of the signal, where the weights come from the filter function evaluated at different time positions. Geometrically, this involves reflecting one function, shifting it along the x-axis, and measuring the overlapping area—a process that captures how the signal and filter interact at each point.

Common Pitfalls and Tips

Master convolution by avoiding these frequent mistakes and applying these practical insights.

  1. Index ordering matters in computation — When manually calculating convolution terms, ensure that as one sequence index increases, the other decreases. Reversing this pairing produces incorrect results. Many errors arise from accidentally summing a₀×b₀ + a₁×b₁ instead of the correct a₀×bₙ + a₁×bₙ₋₁ pattern.
  2. Length of the output sequence — The convolution of a sequence with <em>m</em> elements and a sequence with <em>n</em> elements produces <em>m</em> + <em>n</em> − 1 output elements. Forgetting this rule leads to incomplete results or confusion about how many terms to calculate.
  3. Commutativity simplifies verification — Since convolution is commutative (f ∗ g = g ∗ f), you can compute the operation using either sequence as the 'filter' and the other as the 'signal'. If your calculation seems off, try reversing the sequence order as a sanity check.
  4. Zero-padding affects interpretation — When working with real applications, padding input sequences with zeros changes the convolution output. In signal processing, zero-padding extends the signal length; in image processing, it determines how edges are handled. Always be explicit about padding assumptions when comparing results.

Frequently Asked Questions

Why is convolution used in signal processing?

Convolution models how filters modify signals. When you apply a filter (like a blur or edge detector) to an audio track or image, convolution mathematically describes the transformation. The filter function defines the response characteristics—how strongly different frequency components are amplified or attenuated. This makes convolution invaluable for designing audio equalizers, image smoothing algorithms, and noise reduction systems. Understanding convolution helps engineers predict exactly how a filter will affect their data before implementation.

Does the order of sequences matter in convolution?

No, convolution is commutative, meaning the order is irrelevant: <em>f ∗ g = g ∗ f</em>. You'll get identical results regardless of which sequence you designate as the signal or filter. However, order becomes significant when combining multiple convolutions—convolution is also associative, so <em>f ∗ (g ∗ h) = (f ∗ g) ∗ h</em>. Additionally, convolution is distributive over addition: <em>f ∗ (g + h) = (f ∗ g) + (f ∗ h)</em>. These algebraic properties make convolution predictable and composable.

What is the identity element for convolution?

The identity element is the unit impulse sequence: {1, 0, 0, 0, ...}. Also called the Dirac delta impulse or discrete-time impulse, convolving any sequence with this identity produces the original sequence unchanged. This property is why impulse responses are so important in signal processing—they characterize a system's complete behavior. Knowing a system's impulse response lets you predict its output for any input by computing convolution.

How does convolution relate to probability distributions?

In probability theory, a key theorem states that if two independent random variables have probability density functions (PDFs), then the PDF of their sum equals the convolution of their individual PDFs. This connection explains why convolution appears throughout statistics: when you add random variables, their combined behavior is governed by convolution. This principle applies to noise in measurements, sampling distributions, and many other stochastic processes.

What's the difference between convolution and correlation?

While superficially similar, convolution and correlation serve different purposes. Convolution reverses one sequence before combining (it involves the term <em>g(x − t)</em>), whereas correlation directly aligns sequences without reversal. In practice, convolution models how systems filter signals, while correlation measures similarity or detects patterns. Both operations use similar computational approaches, but their mathematical definitions and applications differ fundamentally.

Can I use convolution for image processing?

Absolutely. In image processing, convolution applies 2D filters (called kernels) to images. A blur kernel smooths pixels by averaging neighbors; an edge-detection kernel highlights boundaries; a sharpening kernel enhances details. Each pixel in the output image is computed by convolving a small region of the input image with the kernel. This technique powers most image filtering in photography software, medical imaging, and computer vision applications. The same principles that govern 1D signal convolution extend naturally to 2D images.

More math calculators (see all)

Cuboid Surface Area Calculator Tan Inverse Calculator Dividing Exponents Calculator Dodecagon Calculator Fermat's Little Theorem Calculator Greatest Common Denominator Calculator Triangular Numbers Calculator Pyramid Angle Calculator