Skip to main content

Creating an ImageSeries

If you want to use images you have captured with a microscope, rather than one of our public datasets, this guide will walk you through the process of creating an ImageSeries object using fpm-py. The ImageSeries is a collection of low-resolution images captured at different illumination angles, along with their corresponding k-space vectors. This data structure is crucial for performing high-resolution image reconstruction using the Fourier Ptychography algorithm.

Prerequisites​

Before you begin, ensure you have the following:

  • A list of low-resolution images captured at different illumination angles.
  • A corresponding list of k-space vectors, one for each image.
  • The optical magnification used during image capture.
  • The physical pixel size of your imaging sensor.
info

Right now, you need to handle calculating the k-vectors yourself. Future releases of fpm-py will expose a function to make that process easier

Import Necessary Libraries​

Start by importing the required libraries:

import torch
import fpm_py as fpm

Prepare Your Data​

Ensure your low-resolution images and k-space vectors are organized in a list. Each image should be a 2D torch.Tensor, and each k-space vector should be a 1x2 torch.Tensor.

# Example list of low-resolution images
images = [
    torch.rand((256, 256)),  # Replace with actual image
    torch.rand((256, 256)),
    torch.rand((256, 256))
]

# Example list of k-space vectors
k_vectors = [
    torch.tensor([0.1, 0.1]),  # Replace with actual k-vectors
    torch.tensor([0.1, 0.2]),
    torch.tensor([0.2, 0.1])
]
tip

If you are using PIL or numpy to handle your images in python, you need to convert them into pytorch Tensors first. https://stackoverflow.com/questions/69738523/how-to-convert-a-tif-pil-image-to-a-torch-tensor

Create ImageCapture Objects​

For each image and its corresponding k-space vector, create an ImageCapture object. This step pairs the image with its k-space vector, making it ready for inclusion in an ImageSeries.

image_captures = []
for image, k_vector in zip(images, k_vectors):
    capture = fpm.ImageCapture(image=image, k_vector=k_vector)
    image_captures.append(capture)

Define Optical Magnification and Pixel Size​

You need to define the optical magnification and the physical pixel size of your imaging setup. These parameters are necessary for accurate reconstruction.

optical_magnification = 20.0  # Example magnification
pixel_size = 1.12  # Example pixel size in micrometers

Create an ImageSeries Object​

Now, create the ImageSeries object by passing the list of ImageCapture objects, along with the optical magnification and pixel size, to the ImageSeries constructor.

image_series = fpm.ImageSeries(
    image_stack=image_captures,
    optical_magnification=optical_magnification,
    pixel_size=pixel_size
)

Save the ImageSeries (Optional)​

If you wish to save the ImageSeries object to disk for later use, you can use the save() method.

image_series.save("path/to/save/image_series.pt")

Load an ImageSeries from Disk (Optional)​

If you have previously saved an ImageSeries, you can load it back into your workspace using the load() method.

loaded_image_series = fpm.ImageSeries.load("path/to/save/image_series.pt")

Summary​

You have successfully created an ImageSeries object in fpm-py using your low-resolution images and their k-space vectors. This ImageSeries can now be used as input to the reconstruct() function for high-resolution image reconstruction. By organizing your data into ImageCapture objects and grouping them into an ImageSeries, you can easily manage and process your datasets with Fourier Ptychography.

For more information on how to use the reconstruct() function, refer to the next section of the user guide.