fixed reading TIFF real valued intensities in range 0-1 as integer 0 #155
Conversation
Apply Sweep Rules to your PR?
|
src/nyx/image_loader1x.cpp
Outdated
| @@ -40,7 +41,10 @@ bool ImageLoader1x::open(const std::string& fpath) | |||
| else | |||
| { | |||
| if (Nyxus::check_tile_status(fpath)) | |||
| FL = std::make_unique<NyxusGrayscaleTiffTileLoader<uint32_t>> (n_threads, fpath); | |||
| FL = std::make_unique<NyxusGrayscaleTiffTileLoader<uint32_t>> (n_threads, fpath, | |||
There was a problem hiding this comment.
Why are we passing in globals this way instead of modifying the ImageLoader1x constructor?
| @@ -165,10 +174,10 @@ class NyxusGrayscaleTiffTileLoader : public AbstractTileLoader<DataType> | |||
| case 8: | |||
| case 16: | |||
| case 32: | |||
| loadTile<float>(tiffTile, tileDataVec); | |||
| loadTile_real_intens <float> (tiffTile, tileDataVec); | |||
There was a problem hiding this comment.
can we do something like
if (image_options_are_present) {
loadTile_real_intens(...)
} else {
loadTile(...)
}
That way, if we are not dealing with already scaled image (which is not the common case, we don't spend tile inspecting each pixel value while copying.
There was a problem hiding this comment.
Floating point image options (in the form of expected minimum, maximum, and equivalent integer range) are always present, aren't they? And doing your suggested way, we'd need to do a separate pre-scan through an image to figure out its intensity range, that is also a time expense.
| @@ -41,7 +44,12 @@ bool ImageLoader::open(const std::string& int_fpath, const std::string& seg_fpat | |||
| { | |||
| if (Nyxus::check_tile_status(int_fpath)) | |||
| { | |||
| intFL = new NyxusGrayscaleTiffTileLoader<uint32_t>(n_threads, int_fpath); | |||
| intFL = new NyxusGrayscaleTiffTileLoader<uint32_t> ( | |||
There was a problem hiding this comment.
Instead of using globals here, can we modify the open() to pass the fimageOptions?
There was a problem hiding this comment.
My motive was to access the global settings as close to the point of use as possible, as just one control flow branch of many passes through the floating point pixels/voxels manipulation. Otherwise we would need to prepare those global settings to pass as parameters early amid the high-level code of phase 2 (trivials) and 3 (non-trivials) image processing logic where we are not concerned about byte layouts in pixels, and do it twice - in function processDataset() and processNontrivialRois(). Your opinion?
No description provided.