Homomorphic filtering is a signal processing technique that decomposes a signal into convolutional or multiplicative components by mapping it into a domain where these operations become additive. The method applies three steps: (1) take the logarithm of the signal converting multiplication to addition and convolution to addition via the convolution theorem, (2) apply a linear filter in the transformed domain, (3) take the exponential to return to the original domain. The technique enables separation of source and channel effects for example in speech processing it separates vocal excitation from vocal tract filtering, and in image processing it separates illumination from reflectance. The mathematical foundation rests on the homomorphism property of the logarithm: log(a times b) equals log(a) plus log(b). The technique persists through its presence in standard DSP textbooks, implementation in libraries, and continued application in speech analysis, image restoration, and radar. [formal: filtrum homomorphum | substrate: behavior | horizon: generations | explicit: yes | epoch: 0.67]
Accepted ontology entry
homomorphic-filtering
Homomorphic filtering is a signal processing technique that decomposes a signal into convolutional or multiplicative components by mapping it into a domain where these operations become additive. The method applies three steps: (1) take th…
Definition
Why it is in scope
A human-made signal processing technique for separating convolutional and multiplicative components in signals. Introduced by Oppenheim and Sheiner 1967, it transforms signals into a domain where convolution becomes addition via logarithm, applies a linear filter, then transforms back. Persisted through textbook treatment in digital signal processing, implementation in deconvolution, image enhancement, speech processing, and radar signal separation.
Names and aliases
- homomorphic-filteringen · CANONICAL
Relations from this entry
- cmrwraj00012csoaclnxvq1ekDEPENDS_ON →
Homomorphic filtering operates by first taking the logarithm of the signal (converting convolution into addition), then FFT, then filtering in frequency domain, then IFFT, then exponential. The logarithmic transform is the indispensable first step — remove it and the technique cannot convert multiplicative convolution into additive operations. This is a present-tense operational dependency, not historical association.
- cmsp49ekv0413jlssad11emypINSTANCE_OF →
Homomorphic filtering is a specific filtering technique that applies filtering operations in a transformed (log) domain to separate multiplicative components. It is a specific kind of filter, not a separate category. A competent speaker would call it a filter.
- cmspqokmf0684jlss24yd0c6fDERIVED_FROM →
Homomorphic filtering applies cepstral analysis techniques to separate multiplicative components in a signal. The method derives from cepstral analysis — the homomorphic signal is computed via cepstral-domain processing (FFT→log→inverse FFT). Historically, cepstral analysis came first; homomorphic filtering was developed as a specialized application. Which-came-first: cepstral analysis → homomorphic filtering.
- cmsp49ekv0413jlssad11emypDEPENDS_ON →
Homomorphic filtering applies filter operations in a transformed domain — it uses linear filters after applying logarithmic compression to the signal. Remove filters and homomorphic filtering has no operational mechanism. The removal test passes: without filters, homomorphic filtering cannot operate.
Relations to this entry
- cmspjfefp05lxjlssg1w6gyuv← DERIVED_FROM
Homomorphic filtering (the log-FFT-iFFT technique) existed first and produced the cepstrum as its output representation. The quefrency-domain concept was derived from the homomorphic filtering pipeline — ask which existed first: homomorphic filtering predates and feeds into the cepstrum.
- cmsqjg9e8003rnqh9rlrvkyjt← DERIVED_FROM
Liftering derives from homomorphic-filtering: liftering is the quefrency-domain windowing operation that forms the filtering step in homomorphic filtering pipelines. Homomorphic filtering (transform→filter→inverse-transform) existed first as a framework, and liftering is the specific quefrency-domain technique derived from it.
- cmsqg5o3d005o3e320t5a9kgk← DERIVED_FROM
The cepstral-lifter derives from homomorphic filtering. Homomorphic filtering (the broader technique of transform→filter→inverse-transform) was developed in the 1960s. The cepstral-lifter is a specialized application of the homomorphic filtering pipeline applied specifically in the quefrency (cepstral) domain for speech processing. The broader technique existed first and fed into the specialized cepstral variant.
Record identity
- Created
- Aug 12, 2026, 11:48 AM UTC
- Content hash
- 54f0a99042b9f0fd85e9091cb6f5dd72e9ffd8b5c25f0c228184e6a13e36f05b