Correction for non-rigid movement artefacts in calcium imaging using local-global optical flow and PCA-based templates
Research output: Contribution to journal › Conference abstract in journal › Research › peer-review
Objectives: Correction for lateral displacements of the imaged area is often a necessary first step of processing calcium imaging data, especially in awake animal studies. We address two problems: (1) image displacements (warps) can be poorly described by simple rigid-body translations or shifts and can be non-uniform across image; (2) due to fluorescence intensity changes single template image may not be optimal for a subset of the movie frames.
Methods: We address the first problem by using either a combined local/global algorithm of optical flow estimation or an original algorithm based on calculation of optical flow in image patches with global regularization. Both algorithms estimate smooth optical flow fields between a current image and a template image and allow for correction of large-scale displacements by employing a multiscale pyramidal approach. The second problem is solved by using a set of template images, obtained from clusters of image frames in low-dimensional PCA-based space. To allow for efficient storage of the estimated image warps, they can be represented as low-pass DCT coefficients or by other dictionary-based methods.
Conclusions: The proposed pipeline for motion correction of calcium timelapse imaging data is accurate, can represent non-rigid image distortions, robust to noisy data and allows for fast registration of large videos. The implementation is open-source and is programmed in Python, which provides for easy access and merging into downstream image processing workflows.
Methods: We address the first problem by using either a combined local/global algorithm of optical flow estimation or an original algorithm based on calculation of optical flow in image patches with global regularization. Both algorithms estimate smooth optical flow fields between a current image and a template image and allow for correction of large-scale displacements by employing a multiscale pyramidal approach. The second problem is solved by using a set of template images, obtained from clusters of image frames in low-dimensional PCA-based space. To allow for efficient storage of the estimated image warps, they can be represented as low-pass DCT coefficients or by other dictionary-based methods.
Conclusions: The proposed pipeline for motion correction of calcium timelapse imaging data is accurate, can represent non-rigid image distortions, robust to noisy data and allows for fast registration of large videos. The implementation is open-source and is programmed in Python, which provides for easy access and merging into downstream image processing workflows.
Original language | English |
---|---|
Article number | PS04-077 |
Journal | Journal of Cerebral Blood Flow and Metabolism |
Volume | 37 |
Issue number | S1 |
Pages (from-to) | 360-361 |
ISSN | 0271-678X |
DOIs | |
Publication status | Published - Apr 2017 |
ID: 182544693