Temporal Smoothing and Optical Flow Analysis

This repository contains Python scripts for processing image sequences by applying temporal smoothing and computing optical flow using the Lucas-Kanade method.

Features

1. Temporal Smoothing:
   Smoothens a sequence of image frames using a weighted kernel, enhancing temporal continuity while reducing noise. The smoothed frames are saved as .npy files.

2. Optical Flow Calculation:
   Computes the optical flow between consecutive smoothed frames using the Lucas-Kanade method. Outputs include velocity fields (vx, vy) and reliability scores, saved as .npy files.

Installation

Ensure you have the following Python libraries installed:

• numpy
• scipy
• matplotlib
• skimage
• numba
• tifffile
• python-bioformats
• javabridge

You can install the required dependencies via pip if needed.

Usage

1. Temporal Smoothing

This function smooths a sequence of raw image frames. It uses a weighted kernel to smooth frames temporally and stores the smoothed frames as .npy files.

• Inputs:
  • data_folder: Path to the folder containing raw image frames (one timeframe per file).
  • single_timelapse_file: Path to a single file containing all timeframes (for example .czi, .lif, .ome.tif, etc.).
  • new_path: Path where smoothed .npy files will be stored.

The smoothing script now supports two modes:

1. Folder mode: Use data_folder (and optional file_list) when timeframes are separate files.
2. Single-file mode: Use single_timelapse_file when the full timelapse is in one file.

If input files are not TIFF, the script converts frames to TIFF before smoothing. Common formats (for example PNG/JPG) are loaded directly, and Bio-Formats is used as a fallback for unsupported microscopy formats.

Command-line examples

Folder mode (one timeframe per file):

python smoothing.py --new-path E:/Spandan/TTX/TTX_Experiments_08_22/smooth --data-folder E:/Spandan/TTX/TTX_Experiments_08_22/p2 --jump 1

Single-file mode (whole timelapse in one file):

python smoothing.py --new-path E:/Spandan/TTX/TTX_Experiments_08_22/smooth --single-timelapse-file E:/Spandan/TTX/TTX_Experiments_08_22/timelapse.czi --jump 1

2. Optical Flow Calculation

This function calculates the optical flow between consecutive smoothed frames using the Lucas-Kanade method. It computes the x and y velocity components (vx, vy) and reliability scores for each pixel, saving these as .npy files.

• Inputs:
  • smooth_address: Path to the folder containing the smoothed .npy files.
  • save_path: Path where the optical flow results will be saved.
  • sigma: The spread of Gaussian weights used in the optical flow calculation.
  • threshold: The reliability score threshold for valid velocity fields.

Outputs

• Temporal Smoothing: The smoothed frames are saved as .npy files in the new_path directory.
• Optical Flow: The computed optical flow velocity components (vx, vy) and reliability scores are saved as .npz files in the save_path directory.

Example Workflow

1. Step 1: Temporal smoothing of raw image frames is done by calling the smoothing function with the appropriate paths.
2. Step 2: After smoothing, optical flow is computed by calling the optical flow function on the smoothed frames.

Contributing

Feel free to fork the repository and create a pull request for any improvements or bug fixes. Please ensure that any contributions are well-documented and come with the appropriate tests.