Previous | Next --- Slide 51 of 61
Back to Lecture Thumbnails
ngandhi

What exactly does it mean to perform Fourier decomposition on an image? What about an image has "frequencies" that we can determine or split up?

emmaloool

It seems like a Fourier decomposition is a technique useful for computer vision (image processing being a subset). I'm curious to see how it can be useful for graphics applications: how does its use case differ between computer vision and computer graphics, which can be seen as fields answering opposite questions (i.e., vision=pictures->abstract representations, graphics=abstract representations->pictures)? Is there a notion of an "inverse"?

rgrao

@ngandhi, any image can be decomposed into low, medium and high frequency images. Low frequency images have slowly varying components like patches of similar color/texture, high frequencies have fast varying components like edges and noise. So usually doing a "blurring" operation on an image with a box filter or a Gaussian filter will "filter" out the high frequency or noise components, resulting in an image with primarily low-frequency components. On the other hand, it is also possible to "sharpen" an image by enhancing these high-frequency edges or noise. You could look it up on Wikipedia, but this is a general explanation :)