def calculateOpticalFlowsForDataset(self, image_reference, dataset):
# optical flows will be stored here
list_optical_flows = []
# add zero optical flow for the reference image which is at first position
shape_image_reference = image_reference.shape
shape_optical_flow = [shape_image_reference[0],
shape_image_reference[1],
2]
zero_optical_flow = np.zeros(shape_optical_flow, np.float32)
list_optical_flows.append(zero_optical_flow)
# iterate through the dataset and calculate the optical flow for each
# except the first one
num_images = dataset.getImageCount()
for index in range(1, num_images):
print ("calculating optical flow for image ", index)
# Get the image at the index
data = dataset.getData(index)
hdu_image = data["hdu_list"][self.extension].data
image = CommonFunctions.preprocessHduImage(hdu_image,
self.scale_factor)
# @todo: here, do not use config but simply check if matrix is None
# apply the transformation to the input image
if self.config["Processing_Options"]["align_images"] == "True":
# get the image dimension
image_shape = image.shape
# Transform the Image
image = cv2.warpAffine(image,
data["transform_matrix"],
(image_shape[1],image_shape[0]),
flags=cv2.INTER_CUBIC + cv2.WARP_INVERSE_MAP)
# calculate the optical flow (backwards for warping!)
optical_flow = cv2.calcOpticalFlowFarneback(image_reference,
image,
None,
self.pyr_scale,
self.levels,
self.winsize,
self.iterations,
self.poly_n,
self.poly_sigma,
cv2.OPTFLOW_FARNEBACK_GAUSSIAN)
# Write out optical flow images for user evaluation
self.writeOpticalFlowImage(index, optical_flow)
list_optical_flows.append(optical_flow)
return list_optical_flows
## Average a list of optical flows
# @param list_optical_flows list object containing optical flows as numpy arrays
# @return averaged optical flow as numpy array
评论列表
文章目录
