httomolibgpu 5.6__tar.gz → 5.7__tar.gz

{httomolibgpu-5.6/httomolibgpu.egg-info → httomolibgpu-5.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: httomolibgpu
-Version: 5.6
+Version: 5.7
 Summary: Commonly used tomography data processing methods at DLS.
 Author-email: Daniil Kazantsev <daniil.kazantsev@diamond.ac.uk>, Yousef Moazzam <yousef.moazzam@diamond.ac.uk>, Naman Gera <naman.gera@diamond.ac.uk>
 License: BSD-3-Clause

{httomolibgpu-5.6 → httomolibgpu-5.7}/httomolibgpu/misc/denoise.py

@@ -175,7 +175,9 @@ def total_variation_PD(
     __check_variable_type(iterations, [int], "iterations", [], methods_name)
     __check_variable_type(isotropic, [bool], "isotropic", [], methods_name)
     __check_variable_type(nonnegativity, [bool], "nonnegativity", [], methods_name)
-    __check_variable_type(lipschitz_const, [float, int], "lipschitz_const", [], methods_name)
+    __check_variable_type(
+        lipschitz_const, [float, int], "lipschitz_const", [], methods_name
+    )
     __check_variable_type(gpu_id, [int], "gpu_id", [], methods_name)
     __check_variable_type(half_precision, [bool], "half_precision", [], methods_name)
     ###################################

{httomolibgpu-5.6 → httomolibgpu-5.7}/httomolibgpu/misc/morph.py

@@ -80,45 +80,76 @@ def sino_360_to_180(
     )
     __check_variable_type(overlap, [int, float], "overlap", [], methods_name)
     __check_variable_type(side, [str], "side", ["left", "right"], methods_name)
-    __check_if_positive_nonzero(
-        overlap, "overlap", positive=True, nonzero=True, methods_name=methods_name
-    )
     ###################################
     #
     dx, dy, dz = data.shape
 
     overlap = int(np.round(overlap))
-    if overlap >= dz - 1:
-        raise ValueError("Overlap must be less than size of the horizontal detector")
     if overlap % 2 != 0:
         overlap += 1
-
+    overlap_pos = abs(overlap)
+    if overlap_pos >= dz - 1:
+        raise ValueError("Overlap must be less than size of the horizontal detector")
     n = dx // 2
     out = cp.empty((n, dy, 2 * dz - overlap), dtype=data.dtype)
 
     if side == "left":
-        weights = cp.linspace(0, 1.0, overlap, dtype=cp.float32)
-        out[:, :, -dz + overlap :] = data[:n, :, overlap:]
-        out[:, :, : dz - overlap] = data[n : 2 * n, :, overlap:][:, :, ::-1]
-        out[:, :, dz - overlap : dz] = (
-            weights * data[:n, :, :overlap]
-            + (weights * data[n : 2 * n, :, :overlap])[:, :, ::-1]
-        )
+        if overlap == 0:
+            out[:, :, -dz:] = data[:n, :, :]
+            out[:, :, :dz] = data[n : 2 * n, :, :][:, :, ::-1]
+            return out
+        elif overlap < 0:
+            out[:, :, -dz:] = data[:n, :, :]
+            out[:, :, : (dz + overlap_pos)] = cp.pad(
+                data[n : 2 * n, :, :][:, :, ::-1],
+                pad_width=(
+                    (0, 0),
+                    (0, 0),
+                    (0, overlap_pos),
+                ),
+                mode="edge",
+            )
+            return out
+        else:
+            weights = cp.linspace(0, 1.0, overlap, dtype=cp.float32)
+            out[:, :, -dz + overlap :] = data[:n, :, overlap:]
+            out[:, :, : dz - overlap] = data[n : 2 * n, :, overlap:][:, :, ::-1]
+            out[:, :, dz - overlap : dz] = (
+                weights * data[:n, :, :overlap]
+                + (weights * data[n : 2 * n, :, :overlap])[:, :, ::-1]
+            )
     if side == "right":
-        weights = cp.linspace(1.0, 0, overlap, dtype=cp.float32)
-        out[:, :, : dz - overlap] = data[:n, :, :-overlap]
-        out[:, :, -dz + overlap :] = data[n : 2 * n, :, :-overlap][:, :, ::-1]
-        out[:, :, dz - overlap : dz] = (
-            weights * data[:n, :, -overlap:]
-            + (weights * data[n : 2 * n, :, -overlap:])[:, :, ::-1]
-        )
+        if overlap == 0:
+            out[:, :, :dz] = data[:n, :, :]
+            out[:, :, dz::] = data[n : 2 * n, :, :][:, :, ::-1]
+            return out
+        elif overlap < 0:
+            out[:, :, : (dz + overlap_pos)] = cp.pad(
+                data[:n, :, :],
+                pad_width=(
+                    (0, 0),
+                    (0, 0),
+                    (0, overlap_pos),
+                ),
+                mode="edge",
+            )
+            out[:, :, (dz + overlap_pos) : :] = data[n : 2 * n, :, :][:, :, ::-1]
+            return out
+        else:
+            weights = cp.linspace(1.0, 0, overlap, dtype=cp.float32)
+            out[:, :, : dz - overlap] = data[:n, :, :-overlap]
+            out[:, :, -dz + overlap :] = data[n : 2 * n, :, :-overlap][:, :, ::-1]
+            out[:, :, dz - overlap : dz] = (
+                weights * data[:n, :, -overlap:]
+                + (weights * data[n : 2 * n, :, -overlap:])[:, :, ::-1]
+            )
 
     return cp.pad(
         out,
         pad_width=(
             (0, 0),
             (0, 0),
-            (overlap // 2, overlap // 2),
+            (overlap_pos // 2, overlap_pos // 2),
         ),
         mode="edge",
     )

{httomolibgpu-5.6 → httomolibgpu-5.7}/httomolibgpu/recon/algorithm.py

@@ -56,6 +56,7 @@ __all__ = [
     "CGLS3d_tomobar",
     "FISTA3d_tomobar",
     "ADMM3d_tomobar",
+    "OSEM3d_tomobar",
 ]
 
 input_data_axis_labels = ["angles", "detY", "detX"]  # set the labels of the input data
@@ -126,7 +127,18 @@ def FBP2d_astra(
         filter_type,
         [str],
         "filter_type",
-        ["none", "ram-lak", "shepp-logan", "tukey", "gaussian", "blackman", "kaiser"],
+        [
+            "none",
+            "ram-lak",
+            "cosine",
+            "shepp-logan",
+            "tukey",
+            "gaussian",
+            "blackman",
+            "kaiser",
+            "lanczos",
+            "hann",
+        ],
         methods_name,
     )
     __check_variable_type(
@@ -552,7 +564,7 @@ def FISTA3d_tomobar(
     subsets_number: int
         The number of the ordered subsets to accelerate convergence. Keep the value bellow 10 to avoid divergence.
     data_fidelity: str
-        Data fidelity given as 'LS' (Least Squares), 'PWLS' (Penalised Weightes LS).
+        Data fidelity given as 'LS' (Least Squares)
     regularisation_type: str
         A method to use for regularisation. Currently PD_TV and ROF_TV are available.
     regularisation_parameter: float
@@ -644,7 +656,7 @@ def ADMM3d_tomobar(
     initialisation: Literal["FBP", "CGLS", "SIRT", None] = "FBP",
     ADMM_rho_const: float = 1.0,
     ADMM_relax_par: float = 1.7,
-    regularisation_type: str = "PD_TV",
+    regularisation_type: Literal["ROF_TV", "PD_TV"] = "PD_TV",
     regularisation_parameter: float = 0.0025,
     regularisation_iterations: int = 40,
     regularisation_half_precision: bool = True,
@@ -679,7 +691,7 @@ def ADMM3d_tomobar(
     subsets_number: int
         The number of the ordered subsets to accelerate convergence. The recommended range is between 12 to 24.
     data_fidelity: str
-        Data fidelity given as 'LS' (Least Squares), 'PWLS' (Penalised Weightes LS).
+        Data fidelity given as 'LS' (Least Squares)
     initialisation: str, optional
         Initialise ADMM with the reconstructed image to reduce the number of iterations and accelerate. Choose between 'CGLS' or 'SIRT' when data
         is noisy and/or undersampled. Choose 'FBP' when the data is of better quality (default) or None.
@@ -835,6 +847,125 @@ def ADMM3d_tomobar(
     return cp.require(cp.swapaxes(reconstruction, 0, 1), requirements="C")
 
 
+## %%%%%%%%%%%%%%%%%%%%%%% OSEM reconstruction %%%%%%%%%%%%%%%%%%%%%%%%%%%%  ##
+def OSEM3d_tomobar(
+    data: cp.ndarray,
+    angles: np.ndarray,
+    center: Optional[float] = None,
+    detector_pad: Union[bool, int] = False,
+    recon_size: Optional[int] = None,
+    recon_mask_radius: Optional[float] = 0.95,
+    iterations: int = 20,
+    subsets_number: int = 12,
+    regularisation_type: Literal["ROF_TV", "PD_TV"] = "PD_TV",
+    regularisation_parameter: float = 1.0,
+    regularisation_iterations: int = 30,
+    regularisation_half_precision: bool = True,
+    nonnegativity: bool = True,
+    gpu_id: int = 0,
+) -> cp.ndarray:
+    """
+    Ordered-Subsets Expectation-Maximisation method is the accelerated Maximum Likelihood Expectation-Maximisation (MLEM) algorithm.
+    Can be coupled with various types of regularisation or denoising operations :cite:`kazantsev2019ccpi` (currently accepts ROF_TV and PD_TV regularisations only).
+    Should be applied to reconstruct emission-type measurements, e.g., XRF tomography measurements.
+
+    Parameters
+    ----------
+    data : cp.ndarray
+        Projection data as a CuPy array.
+    angles : np.ndarray
+        An array of angles given in radians.
+    center : float, optional
+        The center of rotation (CoR).
+    detector_pad : bool, int
+        Detector width padding with edge values to remove circle/arc type artifacts in the reconstruction. Set to True to perform
+        an automated padding or specify a certain value as an integer.
+    recon_size : int, optional
+        The squared size of the reconstructed slice. By default (None), the reconstructed size will be the dimension of the horizontal detector.
+    recon_mask_radius: float, optional
+        The radius of the circular mask that applies to the reconstructed slice in order to crop
+        out some undesirable artifacts. The values outside the given diameter will be set to zero.
+        To implement the cropping one can use the range [0.7-1.0] or set to None (2.0) when no cropping is needed.
+    iterations : int
+        The number of OSEM algorithm iterations. For OS method 20 interations is normally sufficient, while for MLEM one should run 300-500 iterations.
+    subsets_number: int
+        The number of the ordered subsets to accelerate convergence. One can set 'subsets_number' to 1 to achieve MLEM, but the number of 'iterations' should be also increased.
+    data_fidelity: str
+        Data fidelity given as 'LS' (Least Squares)
+    regularisation_type: str
+        A method to use for regularisation. Currently PD_TV and ROF_TV are available.
+    regularisation_parameter: float
+        The main regularisation parameter to control the amount of smoothing/noise removal. Larger values lead to stronger smoothing.
+    regularisation_iterations: int
+        The number of iterations for regularisers (aka INNER iterations).
+    regularisation_half_precision: bool
+        Perform faster regularisation computation in half-precision with a very minimal sacrifice in quality.
+    nonnegativity : bool
+        Impose nonnegativity constraint on the reconstructed image.
+    gpu_id : int
+        A GPU device index to perform operation on.
+
+    Returns
+    -------
+    cp.ndarray
+        The OSEM reconstructed volume as a CuPy array.
+    """
+    ### Data and parameters checks ###
+    methods_name = "OSEM3d_tomobar"
+    __common_data_parameters_check(
+        data,
+        angles,
+        methods_name,
+        center,
+        detector_pad,
+        recon_size,
+        recon_mask_radius,
+        gpu_id,
+    )
+    __common_iterative_basic_parameters_check(methods_name, iterations, nonnegativity)
+    __common_iterative_parameters_check(
+        methods_name,
+        subsets_number,
+        regularisation_type,
+        regularisation_parameter,
+        regularisation_iterations,
+        regularisation_half_precision,
+    )
+    ###################################
+
+    RecToolsCP = _instantiate_iterative_recon_class(
+        data,
+        angles,
+        center,
+        detector_pad,
+        recon_size,
+        subsets_number,
+        gpu_id,
+    )
+
+    _data_ = {
+        "projection_data": data,
+        "data_axes_labels_order": input_data_axis_labels,
+    }
+
+    _algorithm_ = {
+        "iterations": iterations,
+        "nonnegativity": nonnegativity,
+        "recon_mask_radius": recon_mask_radius,
+    }
+
+    _regularisation_ = {
+        "method": regularisation_type,  # Selected regularisation method
+        "regul_param": regularisation_parameter,  # Regularisation parameter
+        "iterations": regularisation_iterations,  # The number of regularisation iterations
+        "half_precision": regularisation_half_precision,  # enabling half-precision calculation
+    }
+
+    reconstruction = RecToolsCP.OSEM(_data_, _algorithm_, _regularisation_)
+    cp._default_memory_pool.free_all_blocks()
+    return cp.require(cp.swapaxes(reconstruction, 0, 1), requirements="C")
+
+
 ## %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%  ##
 def _instantiate_direct_recon_class(
     data: cp.ndarray | Tuple[int, int, int],

{httomolibgpu-5.6 → httomolibgpu-5.7/httomolibgpu.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: httomolibgpu
-Version: 5.6
+Version: 5.7
 Summary: Commonly used tomography data processing methods at DLS.
 Author-email: Daniil Kazantsev <daniil.kazantsev@diamond.ac.uk>, Yousef Moazzam <yousef.moazzam@diamond.ac.uk>, Naman Gera <naman.gera@diamond.ac.uk>
 License: BSD-3-Clause