Bugfixing legacy inversion of drift vector

Author: Macl-I

src/napari_fast4dreg/_fast4Dreg_functions.py

@@ -392,7 +392,6 @@ def translate_z_stack(_image, _shift, transform_type='z'):
 def _translate_z_stack_cpu(_image, _shift, transform_type):
     """CPU implementation using scipy.ndimage."""
     _image_out = []
-    _shift = - _shift  # Negate shift for correct direction
     return_numpy = isinstance(_image, np.ndarray) and _image.ndim == 3
     image_iter = [_image] if return_numpy else _image
 
@@ -458,7 +457,6 @@ def _translate_z_stack_cpu(_image, _shift, transform_type):
 def _translate_z_stack_gpu(_image, _shift, transform_type):
     """GPU implementation using pyclesperanto."""
     _image_out = []
-    _shift = - _shift  # Negate shift for correct direction
 
     def _apply_gpu_transform(img_data, shift_params, ttype):
         """Helper to apply GPU transformation to numpy array."""
@@ -835,4 +833,4 @@ def read_tmp_data(_path, _new_shape=None):
 # - This allows independent, efficient access to individual channels and timepoints
 # - Built-in compression (Blosc/LZ4) reduces disk usage by ~2-3x
 # - Better random access patterns compared to NPY stacks
-# - Particularly efficient for multi-channel images 
+# - Particularly efficient for multi-channel images