I'm doing research on the Gabor filter. When I Googled it, I found very long and complicated articles. Can anyone help me to find a simple explanation about it, or recommend a website or article to read?

I want to understand this filter in order to use it in Matlab.

  • $\begingroup$ I just wanted to know how certain parameters like aspect ratio, bandwidth etc effect the outcome of applying a gabor filter to any image for segmentation $\endgroup$
    – vini
    Apr 29, 2012 at 14:28
  • 1
    $\begingroup$ I did my masters project in Gabor filters what do you want exactly? Do you want to concentrate on equations followed in Gabor or half peak magnitude contour of Gabor .i am doing research on this.add me I will help you. $\endgroup$
    – user3827
    Feb 20, 2013 at 19:57
  • $\begingroup$ Can I add u? @user3827 $\endgroup$ Dec 9, 2013 at 15:22
  • $\begingroup$ There is also another type of Gabor filters which consider the log frequency space... Did you have a look at en.wikipedia.org/wiki/Log_Gabor_filter ? $\endgroup$
    – meduz
    Apr 26, 2018 at 19:17

3 Answers 3


Gabor filters are orientation-sensitive filters, used for texture analysis.

The typically travel in packs, one for each direction. A gabor filter set with a given direction gives a strong response for locations of the target images that have structures in this given direction. For instance, if your target image is made of a periodic grating in a diagonal direction, a gabor filter set will give you a strong response only if its direction matches the one of the grating.

I know it is used a lot in character recognition and fingerprint enhancement. I (try to) use it in bio-medical imaging to characterizes the main orientation in fibril structures.

Here is a very good tutorial by Javier Movellan, pdf

And if you can read french, here is pdf on the creation of filter banks by Adrien Marion

  • $\begingroup$ I just saw this, and cant help but think it might be applicable to this problem ? For that problem I can look at the spectrogram as an image - how would you say this 'orientation sensitive' filter would be applicable? $\endgroup$
    – Spacey
    Mar 2, 2012 at 17:56
  • $\begingroup$ @Mohammad: I honestly don't know. For the pb you are mentioning, working in the F x T domain as you did, and maybe segment spots and count how many you have. $\endgroup$
    – Jean-Yves
    Mar 2, 2012 at 19:02
  • $\begingroup$ remark: Gabor filters don't have to be orientation-sensitive. there are degenerate cases that are not. $\endgroup$
    – thang
    Feb 22, 2013 at 14:08
  • 1
    $\begingroup$ they are also used in your retina, apparently. $\endgroup$
    – endolith
    May 6, 2013 at 13:39
  • $\begingroup$ What if I don't care about orientations, is there any method or variations of Gabor that gives the orientation-independent frequency responses? $\endgroup$
    – Jason
    Jun 21, 2019 at 3:13

A Gabor filter is some parametrization of the idea of edges. This combines two somewhat contradictory ideas: an abrupt transition AND some fuzzy idea of where it is localized.

It is mathematically a clever idea as it translates well in the Fourier domain: the Fourier transform of a Gabor is a Gaussian in Fourier space, and a Gaussian blob is the most neutral guess of something blurry you can make (think of throwing darts and looking at the patterns of hits).

Multiresolution schemes.

As a consequence, when you use a Gabor, there is no 'right' formula: it all depends on what visual feature you want to detect. In visual neuroscience, a popular choice is to chose a Gabor that corresponds in Fourier space to a blob on the logarithm of frequencies (as from the Weber law, we are sensitive to relative differences of frequencies). These are log-Gabor filters.

To understand Gabor filters, check first what filter parameterization would be best for your particular application.


It is an edge detector. It just applies the Gabor Transform. The Gabor filter is basically a Gaussian (with variances sx and sy along x and y-axes respectively) modulated by a complex sinusoid (with centre frequencies U and V along x and y-axes respectively). See an example here.

  • $\begingroup$ This is interesting... I am surprised I have not heard of gabor transform before - how is it different or more advantageous than the hilbert transformer however? You say edge-detector, so if I have a sudden sharp increase in energy, this filter can be used to beat down noise but simultaneously preserve the edge?... $\endgroup$
    – Spacey
    Mar 2, 2012 at 17:57
  • $\begingroup$ remark: this is also not strictly true. you've limited your set to only separable Gabor filters. in fact, the Gaussian envelop does not have to have variance along x and y-axes. the major and minor axes can be tilted. $\endgroup$
    – thang
    Feb 22, 2013 at 14:14

Not the answer you're looking for? Browse other questions tagged or ask your own question.