plenoptic.process.LaplacianPyramid

Note

This object is a torch.nn.Module. It therefore has all the methods and attributes from that class, even though they are not documented here (to avoid cluttering this page).

class plenoptic.process.LaplacianPyramid(n_scales=5, scale_filter=False)

Bases: Module

Laplacian Pyramid in Torch.

The Laplacian pyramid (Burt and Adelson, 1983, [1]) is a multiscale image representation. It decomposes the image by computing the local mean using Gaussian blurring filters and subtracting it from the image and repeating this operation on the local mean itself after downsampling. This representation is overcomplete and invertible.

Parameters:

• n_scales (int (default: 5)) – Number of scales to compute.

• scale_filter (bool (default: False)) – If True, the norm of the downsampling/upsampling filter is 1. If False, it is 2. If the norm is 1, the image is multiplied by 4 during the upsampling operation; the net effect is that the \(n\) -th scale of the pyramid is divided by \(2^n\).

n_scales

Number of computed scales.

Type:

int

scale_filter

Whether the filter is scaled or not.

Type:

bool

References

[1]

Burt, P. and Adelson, E., 1983. The Laplacian pyramid as a compact image code. IEEE Transactions on communications, 31(4), pp.532-540.

Examples

>>> import plenoptic as po
>>> lpyr = po.process.LaplacianPyramid(n_scales=4, scale_filter=True)

Methods

forward(x)	Build the Laplacian pyramid of an image.
recon_pyr(y)	Reconstruct the image from its Laplacian pyramid coefficients.

forward(x)

Build the Laplacian pyramid of an image.

Builds a Laplacian pyramid of height self.n_scales. Because the tensor at each scale will have a different height and width, we return a list of tensors instead of a single tensor.

Parameters:

x (Tensor) – Image, or batch of images of shape (batch, channel, height, width). If there are multiple batches or channels, the Laplacian is computed separately for each of them.

Return type:

list[Tensor]

Returns:

y – Laplacian pyramid representation, each element of the list corresponds to a scale, from fine to coarse.

Examples

>>> import plenoptic as po
>>> img = po.data.einstein()
>>> lpyr = po.process.LaplacianPyramid()
>>> po.plot.imshow(lpyr(img))
<PyrFigure ...>

(png, hires.png, pdf)

recon_pyr(y)

Reconstruct the image from its Laplacian pyramid coefficients.

The input to recon_pyr should be list of tensors similar to those returned by self.forward.

Parameters:

y (list[Tensor]) – Laplacian pyramid representation, each element of the list corresponds to a scale, from fine to coarse. len(y) should be self.n_scales.

Return type:

Tensor

Returns:

x – Image, or batch of images.

Examples

>>> import plenoptic as po
>>> import torch
>>> img = po.data.einstein()
>>> lpyr = po.process.LaplacianPyramid()
>>> coeffs = lpyr(img)
>>> recon = lpyr.recon_pyr(coeffs)
>>> torch.allclose(img, recon)
True
>>> titles = ["Original", "Reconstructed", "Difference"]
>>> po.plot.imshow([img, recon, img - recon], title=titles)
<PyrFigure ...>

(png, hires.png, pdf)