Merge pull request #257 from DiamondLightSource/optimize-stripe-removal-radway-87

Optimize `remove_all_stripe` and `remove_stripe_based_sorting`

Author: dkazanc

httomolibgpu/misc/sorting.py

@@ -0,0 +1,54 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ---------------------------------------------------------------------------
+# Copyright 2026 Diamond Light Source Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either ecpress or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ---------------------------------------------------------------------------
+# Created By  : Tomography Team at DLS <scientificsoftware@diamond.ac.uk>
+# Created Date: 22 April 2026
+# ---------------------------------------------------------------------------
+
+from typing import Tuple
+import cupy as cp
+
+
+def argsort_with_reverse(
+    data: cp.ndarray, axis: int = -1
+) -> Tuple[cp.ndarray, cp.ndarray]:
+    """
+    Compute sorting indices for an 1D or 2D array, and efficiently compute the indices to revert the sort.
+    """
+    dim = len(data.shape)
+    if 1 <= dim <= 2:
+        pass
+    else:
+        raise ValueError("only 1D and 2D arrays are supported")
+    if axis < 0:
+        axis = dim + axis
+    if axis >= dim:
+        raise ValueError("invalid axis")
+    sort_indices = cp.argsort(data, axis=axis)
+    reverse_sort_indices = cp.empty_like(sort_indices)
+    if dim == 1:
+        reverse_sort_indices[sort_indices] = cp.arange(0, data.size)
+    elif axis == 0:  # sort rows
+        nrows, ncols = data.shape
+        cols = cp.arange(ncols)[None, :]
+        reverse_sort_indices[sort_indices, cols] = cp.arange(nrows)[:, None]
+    else:  # sort columns
+        nrows, ncols = data.shape
+        rows = cp.arange(nrows)[:, None]
+        reverse_sort_indices[rows, sort_indices] = cp.arange(ncols)
+
+    return sort_indices, reverse_sort_indices

httomolibgpu/prep/stripe.py

@@ -23,6 +23,7 @@
 import numpy as np
 import pywt
 from httomolibgpu import cupywrapper
+from httomolibgpu.misc.sorting import argsort_with_reverse
 
 cp = cupywrapper.cp
 cupy_run = cupywrapper.cupy_run
@@ -116,21 +117,18 @@ def _rs_sort(sinogram, size, dim):
     """
     Remove stripes using the sorting technique.
     """
-    sinogram = cp.transpose(sinogram)
+    #: Sort each row of the sinogram by its grayscale values
+    sortvals, sortvals_reverse = argsort_with_reverse(sinogram, axis=0)
 
-    #: Sort each column of the sinogram by its grayscale values
-    #: Keep track of the sorting indices so we can reverse it below
-    sortvals = cp.argsort(sinogram, axis=1)
-    sortvals_reverse = cp.argsort(sortvals, axis=1)
-    sino_sort = cp.take_along_axis(sinogram, sortvals, axis=1)
+    sino_sort = cp.take_along_axis(sinogram, sortvals, axis=0)
 
     #: Now apply the median filter on the sorted image along each row
-    sino_sort = median_filter(sino_sort, (size, 1) if dim == 1 else (size, size))
+    sino_sort = median_filter(sino_sort, (1, size) if dim == 1 else (size, size))
 
     #: step 3: re-sort the smoothed image columns to the original rows
-    sino_corrected = cp.take_along_axis(sino_sort, sortvals_reverse, axis=1)
+    sino_corrected = cp.take_along_axis(sino_sort, sortvals_reverse, axis=0)
 
-    return cp.transpose(sino_corrected)
+    return sino_corrected
 
 
 def remove_stripe_ti(
@@ -821,6 +819,7 @@ def remove_all_stripe(
     la_size: int = 61,
     sm_size: int = 21,
     dim: Literal[1, 2] = 1,
+    normalize: bool = False,
 ) -> cp.ndarray:
     """
     Remove all types of stripe artifacts from sinogram using Nghia Vo's
@@ -839,6 +838,8 @@ def remove_all_stripe(
         Window size of the median filter to remove small-to-medium stripes.
     dim : {1, 2},
         Dimension of the window.
+    normalize : bool,
+        Controls whether to normalize while removing large stripes
 
     Returns
     -------
@@ -859,10 +860,10 @@ def remove_all_stripe(
     __check_variable_type(dim, [int], "dim", [1, 2], methods_name)
     ###################################
 
-    matindex = _create_matindex(data.shape[2], data.shape[0])
     for m in range(data.shape[1]):
         sino = data[:, m, :]
-        sino = _rs_dead(sino, snr, la_size, matindex)
+        sino = _rs_dead(sino, snr, la_size)
+        sino = _rs_large(sino, snr, la_size, normalize)
         sino = _rs_sort(sino, sm_size, dim)
         sino = cp.nan_to_num(sino)
         data[:, m, :] = sino
@@ -910,15 +911,17 @@ def _detect_stripe(listdata, snr):
     return listmask
 
 
-def _rs_large(sinogram, snr, size, matindex, drop_ratio=0.1, norm=True):
+def _rs_large(sinogram, snr, size, normalize, drop_ratio=0.1):
     """
     Remove large stripes.
     """
     drop_ratio = max(min(drop_ratio, 0.8), 0)  # = cp.clip(drop_ratio, 0.0, 0.8)
     nrow, ncol = sinogram.shape
     ndrop = int(0.5 * drop_ratio * nrow)
-    sinosort = cp.sort(sinogram, axis=0)
+    sort_indices, sort_indices_reverse = argsort_with_reverse(sinogram, axis=0)
+    sinosort = cp.take_along_axis(sinogram, sort_indices, axis=0)
     sinosmooth = median_filter(sinosort, (1, size))
+
     list1 = cp.mean(sinosort[ndrop : nrow - ndrop], axis=0)
     list2 = cp.mean(sinosmooth[ndrop : nrow - ndrop], axis=0)
     listfact = list1 / list2
@@ -927,30 +930,16 @@ def _rs_large(sinogram, snr, size, matindex, drop_ratio=0.1, norm=True):
     listmask = _detect_stripe(listfact, snr)
     listmask = binary_dilation(listmask, iterations=1).astype(listmask.dtype)
     matfact = cp.tile(listfact, (nrow, 1))
-    # Normalize
-    if norm is True:
+    if normalize:
         sinogram = sinogram / matfact
-    sinogram1 = cp.transpose(sinogram)
-    matcombine = cp.asarray(cp.dstack((matindex, sinogram1)))
-
-    ids = cp.argsort(matcombine[:, :, 1], axis=1)
-    matsort = matcombine.copy()
-    matsort[:, :, 0] = cp.take_along_axis(matsort[:, :, 0], ids, axis=1)
-    matsort[:, :, 1] = cp.take_along_axis(matsort[:, :, 1], ids, axis=1)
-
-    matsort[:, :, 1] = cp.transpose(sinosmooth)
-    ids = cp.argsort(matsort[:, :, 0], axis=1)
-    matsortback = matsort.copy()
-    matsortback[:, :, 0] = cp.take_along_axis(matsortback[:, :, 0], ids, axis=1)
-    matsortback[:, :, 1] = cp.take_along_axis(matsortback[:, :, 1], ids, axis=1)
-
-    sino_corrected = cp.transpose(matsortback[:, :, 1])
+        sort_indices, sort_indices_reverse = argsort_with_reverse(sinogram, axis=0)
+    sino_corrected = cp.take_along_axis(sinosmooth, sort_indices_reverse, axis=0)
     listxmiss = cp.where(listmask > 0.0)[0]
     sinogram[:, listxmiss] = sino_corrected[:, listxmiss]
     return sinogram
 
 
-def _rs_dead(sinogram, snr, size, matindex, norm=True):
+def _rs_dead(sinogram, snr, size):
     """remove unresponsive and fluctuating stripes"""
     sinogram = cp.copy(sinogram)  # Make it mutable
     nrow, _ = sinogram.shape
@@ -979,9 +968,6 @@ def _rs_dead(sinogram, snr, size, matindex, norm=True):
             sinogram[:, listx[ids]] - sinogram[:, listx[ids - 1]]
         )
 
-    # Remove residual stripes
-    if norm is True:
-        sinogram = _rs_large(sinogram, snr, size, matindex)
     return sinogram
 
 

pyproject.toml

@@ -60,7 +60,8 @@ dev = [
   "imageio",
   "h5py",
   "pre-commit",
-  "pyfftw"
+  "pyfftw",
+  "tomophantom"
 ]
 
 

tests/test_misc/test_sorting.py

@@ -0,0 +1,144 @@
+import pytest
+import cupy as cp
+
+from httomolibgpu.misc.sorting import argsort_with_reverse
+
+
+class TestArgsort1D:
+    def test_1d_sorted(self):
+        data = cp.array([1, 2, 3, 4, 5])
+        sort_idx, rev_idx = argsort_with_reverse(data)
+        cp.testing.assert_array_equal(sort_idx, cp.array([0, 1, 2, 3, 4]))
+        cp.testing.assert_array_equal(data[sort_idx], data)
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.arange(5))
+
+    def test_1d_reverse_sorted(self):
+        data = cp.array([5, 4, 3, 2, 1])
+        sort_idx, rev_idx = argsort_with_reverse(data)
+        cp.testing.assert_array_equal(sort_idx, cp.array([4, 3, 2, 1, 0]))
+        cp.testing.assert_array_equal(data[sort_idx], cp.array([1, 2, 3, 4, 5]))
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.arange(5))
+
+    def test_1d_random(self):
+        data = cp.array([10, 3, 7, 1, 9])
+        sort_idx, rev_idx = argsort_with_reverse(data)
+        sorted_data = data[sort_idx]
+        assert cp.all(sorted_data[:-1] <= sorted_data[1:])
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.arange(5))
+
+    def test_1d_duplicates(self):
+        data = cp.array([2, 1, 2, 1])
+        sort_idx, rev_idx = argsort_with_reverse(data)
+        sorted_data = data[sort_idx]
+        assert cp.all(sorted_data[:-1] <= sorted_data[1:])
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.arange(4))
+
+    def test_1d_single_element(self):
+        data = cp.array([42])
+        sort_idx, rev_idx = argsort_with_reverse(data)
+        cp.testing.assert_array_equal(sort_idx, cp.array([0]))
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.array([0]))
+
+    def test_1d_negative_axis(self):
+        data = cp.array([3, 1, 2])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=-1)
+        cp.testing.assert_array_equal(sort_idx, cp.argsort(data))
+        cp.testing.assert_array_equal(rev_idx[sort_idx], cp.arange(3))
+
+
+class TestArgsort2D:
+    def test_2d_axis0_sorted(self):
+        data = cp.array([[1, 2], [3, 4], [5, 6]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=0)
+        sorted_data = data[sort_idx, cp.arange(data.shape[1])[None, :]]
+        assert cp.all(sorted_data[:-1, :] <= sorted_data[1:, :])
+        rows, cols = data.shape
+        col_indices = cp.arange(cols)[None, :]
+        cp.testing.assert_array_equal(
+            rev_idx[sort_idx, col_indices], cp.tile(cp.arange(rows)[:, None], [1, 2])
+        )
+
+    def test_2d_axis0_random(self):
+        data = cp.array([[5, 1], [2, 4], [3, 6]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=0)
+        rows, cols = data.shape
+        col_indices = cp.arange(cols)[None, :]
+        cp.testing.assert_array_equal(
+            rev_idx[sort_idx, col_indices], cp.tile(cp.arange(rows)[:, None], [1, 2])
+        )
+
+    def test_2d_axis0_negative_axis(self):
+        data = cp.array([[1, 2], [3, 4]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=-2)
+        rows, cols = data.shape
+        col_indices = cp.arange(cols)[None, :]
+        cp.testing.assert_array_equal(
+            rev_idx[sort_idx, col_indices], cp.tile(cp.arange(rows)[:, None], [1, 2])
+        )
+
+    def test_2d_axis1_sorted(self):
+        data = cp.array([[1, 2, 3], [4, 5, 6]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=1)
+        rows, cols = data.shape
+        row_indices = cp.arange(rows)[:, None]
+        cp.testing.assert_array_equal(
+            rev_idx[row_indices, sort_idx], cp.tile(cp.arange(cols), [2, 1])
+        )
+
+    def test_2d_axis1_random(self):
+        data = cp.array([[3, 1, 2], [6, 4, 5]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=1)
+        rows, cols = data.shape
+        row_indices = cp.arange(rows)[:, None]
+        cp.testing.assert_array_equal(
+            rev_idx[row_indices, sort_idx], cp.tile(cp.arange(cols), [2, 1])
+        )
+
+    def test_2d_axis1_negative_axis(self):
+        data = cp.array([[1, 2], [3, 4]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=-1)
+        rows, cols = data.shape
+        row_indices = cp.arange(rows)[:, None]
+        cp.testing.assert_array_equal(
+            rev_idx[row_indices, sort_idx], cp.tile(cp.arange(cols), [2, 1])
+        )
+
+    def test_2d_single_row(self):
+        data = cp.array([[3, 1, 2]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=1)
+        rows, cols = data.shape
+        row_indices = cp.arange(rows)[:, None]
+        cp.testing.assert_array_equal(
+            rev_idx[row_indices, sort_idx], cp.arange(cols)[None, :]
+        )
+
+    def test_2d_single_col(self):
+        data = cp.array([[3], [1], [2]])
+        sort_idx, rev_idx = argsort_with_reverse(data, axis=0)
+        rows, cols = data.shape
+        col_indices = cp.arange(cols)[None, :]
+        cp.testing.assert_array_equal(
+            rev_idx[sort_idx, col_indices], cp.arange(rows)[:, None]
+        )
+
+
+class TestArgsortErrors:
+    def test_invalid_dim_3d(self):
+        data = cp.ones((2, 2, 2))
+        with pytest.raises(ValueError):
+            argsort_with_reverse(data)
+
+    def test_invalid_dim_0d(self):
+        data = cp.array(1)
+        with pytest.raises(ValueError):
+            argsort_with_reverse(data)
+
+    def test_invalid_axis_1d(self):
+        data = cp.array([1, 2, 3])
+        with pytest.raises(ValueError):
+            argsort_with_reverse(data, axis=1)
+
+    def test_invalid_axis_2d(self):
+        data = cp.array([[1, 2], [3, 4]])
+        with pytest.raises(ValueError):
+            argsort_with_reverse(data, axis=2)