def __getitem__(self, idx):
if self.Dataset_name is "train":
image, image_origin = self.imgset[idx]
else:
image, image_name, image_origin = self.imgset[idx]
# print("Image max {} min {} mean {}".format(np.array(image).max(), np.array(image).min(), np.array(image).mean()))
sinogram_full = self.project(image, self.proj_id_full)
sinogram_sparse = sparse_view_f(sinogram_full,
self.geo_full["sino_views"],
self.geo_sparse["sino_views"])
sinogram_inter = my_map_coordinates(
sinogram_sparse,
(self.geo_full["sino_views"], self.geo_full["nDetecU"]),
order=3)
image_full = self.fbp(sinogram_full, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
image_sparse = self.fbp(sinogram_sparse, self.proj_id_sparse,
self.proj_geom_sparse, self.vol_geom_sparse)
image_inter = self.fbp(sinogram_inter, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
sample = {
"image_full": image_full,
"image_true": image,
"image_origin": image_origin,
"image_sparse": image_sparse,
"image_inter": image_inter,
"sinogram_sparse": sinogram_sparse,
"sinogram_full": sinogram_full
}
return sample
def __getitem__(self, idx):
if self.Dataset_name is "train":
image= self.imgset[idx]
else:
image, image_name = self.imgset[idx]
sinogram_full = self.project(image, self.proj_id_full)
sinogram_sparse = sparse_view_f(sinogram_full, self.geo_full["sino_views"], self.geo_sparse["sino_views"])
image_full = self.fbp(sinogram_full, self.proj_id_full, self.proj_geom_full, self.vol_geom_full)
image_sparse = self.fbp(sinogram_sparse, self.proj_id_sparse, self.proj_geom_sparse, self.vol_geom_sparse)
if self.Dataset_name is "train":
sample = {"image_full": torch.from_numpy(image_full).unsqueeze_(0),
"image_true": image,
"image_sparse": torch.from_numpy(image_sparse).unsqueeze_(0),
"image_res": torch.from_numpy(image_full - image_sparse).unsqueeze_(0)}
else:
sample = {"image_full": torch.from_numpy(image_full).unsqueeze_(0),
"image_true": image,
"image_sparse": torch.from_numpy(image_sparse).unsqueeze_(0),
"image_res": torch.from_numpy(image_full - image_sparse).unsqueeze_(0),
"image_name":image_name}
return sample
def __getitem__(self, idx):
image, image_name = self.imgset[idx]
sinogram_full = self.project(image, self.proj_id)
sinogram_sparse = sparse_view_f(sinogram_full, self.geo_full["sino_views"], self.geo_sparse["sino_views"])
sinogram_inter = my_map_coordinates(sinogram_sparse, (self.geo_full["sino_views"], self.geo_full["nDetecU"]), order=1)
return image, sinogram_full, sinogram_sparse, sinogram_inter, image_name
def __getitem__(self, idx):
if self.Dataset_name is "train":
image = self.imgset[idx]
else:
image, image_name = self.imgset[idx]
# print("Image max {} min {} mean {}".format(np.array(image).max(), np.array(image).min(), np.array(image).mean()))
sinogram_full = self.project(image, self.proj_id_full)
sinogram_sparse = sparse_view_f(sinogram_full,
self.geo_full["sino_views"],
self.geo_sparse["sino_views"])
# sinogram_inter = my_map_coordinates(sinogram_sparse, (self.geo_full["sino_views"], self.geo_full["nDetecU"]), order=1)
# sinogram_inter = my_extension(sinogram_sparse, (self.geo_full["sino_views"], self.geo_full["nDetecU"]))
image_full = self.fbp(sinogram_full, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
image_sparse_0 = self.fbp(sinogram_sparse, self.proj_id_sparse,
self.proj_geom_sparse, self.vol_geom_sparse)
image_i1_pred = self.pred_sample(image_sparse_0, self.model_i1)
sinogram_i1_pred = self.project(image_i1_pred, self.proj_id_full)
sinogram_new = self.updata_sinogram(sinogram_sparse, sinogram_i1_pred)
image_sparse_1 = self.fbp(sinogram_new, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
image_i2_pred = self.pred_sample(image_sparse_1, self.model_i2)
sinogram_i2_pred = self.project(image_i2_pred, self.proj_id_full)
sinogram_new = self.updata_sinogram(sinogram_sparse, sinogram_i2_pred)
image_sparse_2 = self.fbp(sinogram_new, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
image_i3_pred = self.pred_sample(image_sparse_2, self.model_i3)
sinogram_i3_pred = self.project(image_i3_pred, self.proj_id_full)
sinogram_new = self.updata_sinogram(sinogram_sparse, sinogram_i3_pred)
image_sparse_3 = self.fbp(sinogram_new, self.proj_id_full,
self.proj_geom_full, self.vol_geom_full)
if self.Dataset_name is "train":
sample = {
"image_full": torch.from_numpy(image_full).unsqueeze_(0),
"image_true": image,
"image_sparse_0": image_sparse_0,
"image_sparse_1": image_sparse_1,
"image_sparse_2": image_sparse_2,
"image_sparse": torch.from_numpy(image_sparse_3).unsqueeze_(0)
}
else:
sample = {
"image_full": torch.from_numpy(image_full).unsqueeze_(0),
"image_true": image,
"image_sparse_0": image_sparse_0,
"image_sparse_1": image_sparse_1,
"image_sparse_2": image_sparse_2,
"image_sparse": torch.from_numpy(image_sparse_3).unsqueeze_(0),
"image_name": image_name
}
return sample