[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

[gpuCI] Forward-merge branch-22.06 to branch-22.08 [skip gpuci]

This PR fixes a bug in `canny` that was leading to failures in `test_use_quantiles`. It also expands that test case via `pytest.parametrize` to cover additional dtypes

The changes to other test classes are merely refactoring to avoid use of `unittest.TestCase` so that `pytest.parametrize` can be used

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #310

This PR incorporates the two bug fixes from skimage v0.19.3 (released on June 12th) that are relevant to cuCIM.

- bug fix to ensure all color channel share the same histogram bins in `cucim.skimage.exposure.histogram`
- clipping in warp functions should respect user-specified `cval` during clipping even if it is outside the original image range

There was also a fix to Canny edge detection, but we had already resolved that issue in cuCIM. I did go ahead and add the new test case for it to #310.

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #312

With the PR below merged, we no longer set the `CUDAHOSTCXX` variable in any of our CI images. This PR cleans up some references to `CUDAHOSTCXX`.


- rapidsai/gpuci-build-environment#265

Authors:
  - AJ Schmidt (https://github.com/ajschmidt8)

Approvers:
  - Sevag H (https://github.com/sevagh)
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #307

Authors:
   - https://github.com/HesAnEasyCoder
   - Gigon Bae (https://github.com/gigony)

Approvers:
   - Benjamin Zaitlen (https://github.com/quasiben)
   - Gigon Bae (https://github.com/gigony)

replaces #306 having manually fixed the conflict in `CHANGELOG.md`

Authors:
   - Ray Douglass (https://github.com/raydouglass)
   - Gregory Lee (https://github.com/grlee77)

Approvers:
   - Mark Sadang (https://github.com/msadang)

This is an update of the previously-approved #290 from branch-22.06 to branch-22.08

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/gbae
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #339

Update of #278 for branch-22.08 with minor additional fixes. I have tested this locally and it seems to be operating as expected.

Authors:
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/aasthajh

Approvers:
  - https://github.com/jakirkham

URL: #322

This PR expands the approach of #322 to the remaining `cucim.skimage` benchmarks as well

This PR is built on top of a few other open benchmarking PRs, so please review and merge those first: (#322, #323, #290)

Authors:
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/aasthajh

Approvers:
  - https://github.com/jakirkham

URL: #324

Incorporated definition for cuCIM; cuCIM benefits; applicable file formats; and links to blogs, documentation, and webinars.

Authors:
  - https://github.com/HesAnEasyCoder
  - Gigon Bae (https://github.com/gigony)

Approvers:
  - Gigon Bae (https://github.com/gigony)
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/jakirkham

URL: #315

…resholding (#316)

I think we want to allow pixels with substantial attenuation in **any** of the channels rather than requiring substantial attenuation in **all** channels. The Macenko et. al. reference does not actually explain how the threshold was applied (min, max or mean across channels). However, qualitatively, results looked closer to expected with the `any` condition proposed here when I was working on a pending stain normalization example/demo.

I have also seen third party implementations where there is an RGB->LUV conversion and thresholding is done on the L channel rather than in luminance space. That seems reasonable and perhaps is more intuitive in choice of the threshold number (easier to think in fraction of brightness rather than absorbance units). However, making that change would be a bigger deviation from the publication and would involve some small additional overhead of the RGB->luminance conversion.

I have marked this as breaking as it will affect the estimated stain vectors produced by `stain_extraction_pca`.

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #316

### Overview

closes #319

This PR adds an implementation of `scipy.ndimage.distance_transform_edt`. For each foreground pixel in a binary image, this function computes the minimal Euclidean distance to reach a background pixel. This is a SciPy function rather than a scikit-image one so I have put it under `cucim.core` instead of `cucim.skimage`. Longer term we could move this upstream to CuPy, but I think we want to have a copy here so that we can use it in the near term to implement some of the missing scikit-image functionality. This function is used by the following functions in the scikit-image API, so this PR will help us implement these in future PRs:

`skimage.morphology.medial_axis` (**update**: see #342)
`skimage.segmentation.expand_labels` (**update**: see #341)
`skimage.segmentation.chan_vese` (**update**: see #343)
It is also used in examples related to `skimage.segmentation.watershed` and is needed for implementation of ITK's `SignedMaurerDistanceMapImageFilter` for [itk-cucim](InsightSoftwareConsortium/itk_cucim#13).

The algorithm used here is adapted from the C++ code in the [PBA+ repository](https://github.com/orzzzjq/Parallel-Banding-Algorithm-plus) (MIT-licensed)

### extensions made to PBA+ kernel source
- larger sizes can be supported due to runtime code generation of the integer datatypes and encode/decode functions.

### current known limitations
- 2D and 3D only (this should cover the most common use cases)
- The `sampling` argument is not fully supported. This can likely be done with a bit of effort, but will require modifications to the CUDA kernels
- User-specified output arguments are not currently supported. We can potentially add this in the future
- The indices returned by `return_indices=True` are equally valid, but not always identical to those returned by SciPy. This is because there can be multiple indices with an identical distance, so which one gets chosen in that case is implementation-dependent.

### initial benchmarks relative to SciPy

Here `% true` is the percentage of the image that corresponds to foreground pixels. Even for fairly small images there is some benefit, with the benefit becoming two orders of magnitude at larger sizes.

shape      | % true |  cuCIM |  SciPy  | acceleration
-----------------|--------|---------|---------|--------------
(256, 256) | 5 | 0.00108 | 0.00353 | 3.25
(512, 512) | 5 | 0.00108 | 0.01552 | 14.42
(2048, 2048) | 5 | 0.00252 | 0.34434 | 136.86
(4096, 4096) | 5 | 0.00765 | 1.58948 | 207.88
(32, 32, 32) | 5 | 0.00103 | 0.00305 | 2.98
(128, 128, 128) | 5 | 0.00153 | 0.30103 | 196.26
(256, 256, 256) | 5 | 0.00763 | 3.17872 | 416.37
(384, 384, 384) | 5 | 0.02460 | 14.28779 | 580.89
(256, 256) | 50 | 0.00107 | 0.00430 | 4.01
(512, 512) | 50 | 0.00109 | 0.01878 | 17.30
(2048, 2048) | 50 | 0.00299 | 0.39304 | 131.60
(4096, 4096) | 50 | 0.00896 | 1.84686 | 206.19
(32, 32, 32) | 50 | 0.00102 | 0.00361 | 3.53
(128, 128, 128) | 50 | 0.00163 | 0.31657 | 194.66
(256, 256, 256) | 50 | 0.00914 | 3.35914 | 367.49
(384, 384, 384) | 50 | 0.03005 | 13.83219 | 460.30
(256, 256) | 95 | 0.00107 | 0.00344 | 3.22
(512, 512) | 95 | 0.00108 | 0.01638 | 15.17
(2048, 2048) | 95 | 0.00314 | 0.36996 | 117.81
(4096, 4096) | 95 | 0.00943 | 1.90475 | 202.05
(32, 32, 32) | 95 | 0.00102 | 0.00314 | 3.07
(128, 128, 128) | 95 | 0.00180 | 0.28843 | 159.80
(256, 256, 256) | 95 | 0.01073 | 3.23450 | 301.41
(384, 384, 384) | 95 | 0.03577 | 12.40526 | 346.82

### other comments
- can likely reduce memory overhead and improve performance a bit by refactoring some of the pre/post-processing code into elementwise kernels. (e.g. `encode3d`/`decode3d`, etc.)

- (JK) This may be able to leverage CuPy in the future ( cupy/cupy#6919 )

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #318

#328)

In this PR we change the following line:
```Python
conc_norm = conc_raw * normalization_factors[:, cp.newaxis]
```

to instead modify `conc_raw` in-place. The `conc_raw` array is an intermediate array of concentrations, so changing to an in-place operation will not effect user-facing code. The benefit is avoiding an extra copy of the concentrations array. It also avoids an unintended promotion to float64 due to multiplication by a `normalization_factors` array that was potentially higher precision than `conc_raw`.

For a fairly large uint8 input array of shape `(26420, 19920, 3)`, the `conc_raw` array ends up as an ~4GB float32 array. Without the change in this PR there will be a new 8GB float64 `conc_norm` array created.  By avoiding this we substantially reduce memory use. 

We also make sure the `ref_max_conc` and `ref_stain_coeff` inputs to `_normalized_from_concentration` match the precision of the image. Due to this, a 25% improvement in performance was observed (due to float32 rather than float64 computations in the remainder of the `_normalized_from_concentrations` function).

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #328

- Update CHANGELOG.md and README.md for PyPI v22.06.00 release

Authors:
  - Gigon Bae (https://github.com/gigony)
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/jakirkham

URL: #311

…in deprecation warnings (#356)

closes #325

TODO:
- [x] update DeprecationWarning strings

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #356

Allows cuCIM to be installed with CuPy 11.

xref: rapidsai/integration#508

Authors:
  - https://github.com/jakirkham

Approvers:
  - Gregory Lee (https://github.com/grlee77)
  - Sevag H (https://github.com/sevagh)

URL: #357

Incorporating latest blog that dropped 7/27.

Authors:
  - https://github.com/HesAnEasyCoder

Approvers:
  - https://github.com/jakirkham
  - AJ Schmidt (https://github.com/ajschmidt8)

URL: #361

Hopefully addresses some `import` issues seen around modules in this package.

cc @grlee77

Authors:
  - https://github.com/jakirkham

Approvers:
  - Gregory Lee (https://github.com/grlee77)

URL: #359

Fixes similar issue found in: rapidsai/cudf#11400

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - https://github.com/jakirkham

URL: #360

Reverts #357

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - AJ Schmidt (https://github.com/ajschmidt8)
  - https://github.com/jakirkham

URL: #362

closes #334
This adds a simple function that can be used to expand a set of labels in a label image.

It should be reviewed after #318 is merged. The new commits only start from 6241bd2.

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #341

This PR is purely a performance refactor without a change in behavior. The main changes to `morphological_chan_vese` are:

- Use a simplified version of `cupy.gradient` that does not promote all integer types to float64 and omits a couple of division operations by assuming unit spacing. Currently we don't have any places where non-uniform spacing or non first-order edges are needed, but we can always still use `cupy.gradient` for those if they come up in the future. 

- apply kernel fusion via `cupy.fuse()` decorators.

- avoid repeated transfer of footprints to the GPU and reuse the same workspace array for repeated `_curveop` calculations.

Overall, these changes give around 1/3 reduction in run-time for `morphological_chan_vese`.

All other places using `cp.gradient` were also changed to use the simplified `gradient` defined here. These include:
`cucim.skimage.features.hessian_matrix`
`cucim.skimage.features.multiscale_basic_features`
`cucim.skimage.registration.optical_flow_tvl1`

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #340

…ze_colors_pca` (#327)

This PR updates `normalize_colors_pca` to raise a more helpful error message in the case that there is only a single pixel above the specified `beta` threshold. Currently this leads to a not very helpful `DivideByZeroError` because the degrees of freedom during an internal covariance calculation becomes zero in this case.

This does also change the error type when there are no pixels above the threshold from `ValueError` to `RuntimeError` which seems more appropriate to me.

Authors:
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/jakirkham

Approvers:
  - https://github.com/jakirkham

URL: #327

closes #364

This tiny fix resolves the local test failures I saw with CuPy 9.6. I also think the updated code is clearer in any case.

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #365

make sure `cucim.core.morphology` module is included in the core API docs

fix typo in references section of `distance_transform_edt` docstring

setup intersphinx link to SciPy docs (used in `distance_transform_edt`)

add `sphinx.ext.mathjax` for LaTeX rendering

fix sphinx warning/typo related to `image_to_absorbance` docstring

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #367

closes #335

This PR adds a function for segmentation based on active-contours. Test cases are directly from scikit-image, but the code itself was heavily refactored to make extensive use of kernel fusion via `cupy.fuse()`. I will add benchmark results here soon.

It should be reviewed after #318 is merged. The new commits only start from 502f41a

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - https://github.com/jakirkham

URL: #343

### Background

closes #320

This PR adds a fast histogram-based median filter. It is based on a paper by Oded Green [1] as available in [`opencv_contrib`](https://github.com/opencv/opencv_contrib/blob/4.x/modules/cudafilters/src/cuda/median_filter.cu). It is a GPU-based adaptation of an earlier CPU-based publication [2].

This new functionality has been added to the existing `median` function and will be automatically used when possible (although the user can override the automated choice via a new keyword-only `algorithm` parameter.

The CUDA file [here](https://github.com/opencv/opencv_contrib/blob/4.x/modules/cudafilters/src/cuda/median_filter.cu) has a BSD 3-clause license up top, although the repository states that it is Apache 2.0 licensed, so that is likely out of date. We do already list OpenCV's extra modules and the Apache 2.0 license for it in our 3rdparty license folder.

### References

[1] O. Green, "Efficient Scalable Median Filtering Using Histogram-Based Operations," in IEEE Transactions on Image Processing, vol. 27, no. 5, pp. 2217-2228, May 2018, https://doi.org/10.1109/TIP.2017.2781375.

[2] S. Perreault and P. Hebert, "Median Filtering in Constant Time," in IEEE Transactions on Image Processing, vol. 16, no. 9, pp. 2389-2394, Sept. 2007, https://doi.org/10.1109/TIP.2007.902329.

### Improvements

There are a number of generalizations in the implementation here as compared to the one in OpenCV:
- supports odd-sized rectangular footprints as well as square ones
- supports int8, uint16 and int16 in addition to uint8
- the dtype used to store the histogram counts is adapted to the filter's footprint size
- handles many additional boundary extension modes via explicit pre-pad and post-crop (in principle could adapt the kernel for each case to avoid this overhead, but likely not worth the substantial additional effort)
- automatically switches between the sorting-based and histogram-based implementations depending on kernel size (sorting is much faster for very small kernels while the histogram-based approach is much faster for large ones)

### Notes/Limitations

The GPU I tested this on has a static shared memory limit of 48k per block. This means that the `HFine` shared memory array can exceed the available GPU memory once there are more than ~12000 gray levels in the histogram. So, for int16 or uint16, we I can only run the algorithm on images where the min/max range does not exceed that value. One fairly common case in medical imaging is DICOM images with 12-bit range (0-4095) that are stored at 16-bit int. This shared memory limit is compatible with this case.

### Performance

see summary in the next comment

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #317

closes #321 

This PR adds a different implementation of separable convolution that was adapted from `opencv_contrib`. It is not yet ready for review.

Still needs:
- [x] support other boundary modes
- [x] tests
- [x] ~extension to nD~  extension to 3D
- [x] wrappers for existing filters to call this one instead of CuPy's

The key changes in this PR are the new files:
`skimage/filters/_separable_filtering.py`
`skimage/filters/tests/test_separable_filtering.py`

It is based on the approach taken in OpenCV-contrib's [row](https://github.com/opencv/opencv_contrib/blob/4.6.0/modules/cudafilters/src/cuda/row_filter.hpp) and [column]([https://github.com/opencv/opencv_contrib/blob/4.6.0/modules/cudafilters/src/cuda/column_filter.hpp] ) filters, but also supports:
- 3D
- additional dtypes (e.g. complex64)
- all boundary modes from SciPy
- not restricted to kernel size <= 32
- casting behavior to the output matches SciPy rather than OpenCV conventions.
 
A simpler version of the same approach has long been in the CUDA samples [convolutionSeparable.cu example](https://github.com/NVIDIA/cuda-samples/blob/master/Samples/2_Concepts_and_Techniques/convolutionSeparable/convolutionSeparable.cu). The basic idea is:
1.) First stage loads the current patch of the image and its boundaries into shared memory
2.) After synchronization, convolution is performed on the shared memory array.

A lot of CuPy's ndimage code is vendored here for the following reasons:
- We need any `cupyx.scipy.ndimage` functions we call, like `gaussian_filter`, `uniform_filter`, etc. to dispatch to the new convolution implementation when possible.
- the `_get_weights_dtype` utility was changed to promote 8 and 16-bit integers to float32 instead of float64 during convolutions.
- the `_get_output` utility was changed to allocate the output arrays as `empty` rather than `zeros` which is more efficient.
- the `_run_1d_filters` utility was improved so that it avoids an extra copy when the number of filters is even

I will submit PRs for these to CuPy. The `_get_output` change in particular also impacts other morphology and interpolation functions that we use in cuCIM, so I have vendored those here as well. For reference regarding these non-filtering changes, there is modest performance improvement. Grayscale `erosion` on a 4k image became 10% faster and `resize` of a 4k image to HD became 20% faster.

It is not easy to review kernels based on generated code strings, so I have tried to add pretty comprehensive tests covering many kernel sizes, all boundary modes, various dtype combinations, etc.

Authors:
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gigon Bae (https://github.com/gigony)
  - https://github.com/jakirkham

URL: #355

closes #336

This PR adds a function for skeletonization of 2D images via the medial axis transform. 

It should be reviewed after #318 is merged. The new commits only start from 19a6fed.

There is one sequential component to this algorithm that still must be run on the CPU, but the majority of the computations are on the GPU and acceleration is good. I will add benchmark results here soon.

Authors:
  - Gregory Lee (https://github.com/grlee77)
  - https://github.com/jakirkham

Approvers:
  - https://github.com/jakirkham
  - Gigon Bae (https://github.com/gigony)

URL: #342

Add 'tiff.resolution_unit' and 'tiff.x_resolution', 'tiff.y_resolution' metadata to the CuImage object.
(See https://www.awaresystems.be/imaging/tiff/tifftags/resolutionunit.html)

If `resolution_unit` is not 1 (1 = No absolute unit of measurement), `spacing_units` is `['micrometer', 'micrometer', 'color']` and spacing unit is calculated based on `micrometer`.

**spacing_test.py**
```python
from pprint import pprint
from cucim import CuImage
img = CuImage("TCGA-18-3406-01Z-00-DX1_tissue.tif")
pprint(img.metadata, indent=2, compact=True)

```

```bash
❯ python spacing_test.py
{ 'cucim': { 'associated_images': [],
             'channel_names': ['R', 'G', 'B', 'A'],
             'coord_sys': 'LPS',
             'dims': 'YXC',
             'direction': [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]],
             'dtype': {'bits': 8, 'code': 1, 'lanes': 1},
             'ndim': 3,
             'origin': [0.0, 0.0, 0.0],
             'path': '/home/gbae/Downloads/TCGA-18-3406-01Z-00-DX1_tissue.tif',
             'resolutions': { 'level_count': 6,
                              'level_dimensions': [ [16034, 10970],
                                                    [8017, 5485], [4008, 2742],
                                                    [2004, 1371], [1002, 685],
                                                    [501, 342]],
                              'level_downsamples': [ 1.0, 2.0,
                                                     4.000614166259766,
                                                     8.001228332519531,
                                                     16.008296966552734,
                                                     32.040008544921875],
                              'level_tile_sizes': [ [512, 512], [512, 512],
                                                    [512, 512], [512, 512],
                                                    [512, 512], [512, 512]]},
             'shape': [10970, 16034, 4],
             'spacing': [2.0159945487976074, 2.0159945487976074, 1.0],
             'spacing_units': ['micrometer', 'micrometer', 'color'],
             'typestr': '|u1'},
  'tiff': { 'model': '',
            'resolution_unit': 'centimeter',
            'software': '',
            'x_resolution': 4960.3310546875,
            'y_resolution': 4960.3310546875}}
```

**Test install**
```bash
python -m pip install --force-reinstall --extra-index-url https://test.pypi.org/simple/ cucim==0.0.333
```

Address #333 

Signed-off-by: Gigon Bae <gbae@nvidia.com>

Authors:
  - Gigon Bae (https://github.com/gigony)
  - Gregory Lee (https://github.com/grlee77)

Approvers:
  - Gregory Lee (https://github.com/grlee77)

URL: #349

This PR switches the loading of `custom.js` to `defer` because we will need the entire page to be loading until the methods in this script can even execute correctly.

xref: rapidsai/cudf#11465

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - AJ Schmidt (https://github.com/ajschmidt8)

URL: #383

closes #363
by bumping the minimum required scikit-image

Authors:
   - Gregory Lee (https://github.com/grlee77)

Approvers:
   - AJ Schmidt (https://github.com/ajschmidt8)
   - https://github.com/jakirkham