Details preserved unsupervised depth estimation by fusing traditional stereo knowledge from laparoscopic images

Autor: Qingmao Hu, Fucang Jia, Huoling Luo
Rok vydání: 2019
Předmět:
Computer science
Point cloud
Special Issue: Papers from the 13th Workshop on Augmented Environments for Computer Assisted Interventions
heart phantom data
truth depth
Convolutional neural network
medical image processing
principled mask
surgery
unsupervised learning depth estimation approach
Health Information Management
proxy disparity labels
vision-based laparoscope surgical navigation systems
Computer vision
laparoscopic images
Ground truth
unsupervised depth estimation
stereo accuracy
image reconstruction
neighbourhood smoothness term
proxy labels
loss function
lcsh:R855-855.5
unreliable depth measurements
Unsupervised learning
dual encoder-decoder convolutional neural network
traditional stereo method
traditional stereo knowledge
lcsh:Medical technology
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION
phantoms
Health Informatics
Iterative reconstruction
stereo image processing
unsupervised learning
image motion analysis
rectified stereo images
convolutional neural nets
disparity images
parallax effects
ComputingMethodologies_COMPUTERGRAPHICS
constrain neighbouring pixels
Pixel
business.industry
smooth depth surface
confidence measure
Artificial intelligence
partial nephrectomy da vinci surgery dataset
Estimation methods
Parallax
business
hamlyn centre
Zdroj: Healthcare Technology Letters (2019)
Healthcare Technology Letters
ISSN: 2053-3713
Popis: Depth estimation plays an important role in vision-based laparoscope surgical navigation systems. Most learning-based depth estimation methods require ground truth depth or disparity images for training; however, these data are difficult to obtain in laparoscopy. The authors present an unsupervised learning depth estimation approach by fusing traditional stereo knowledge. The traditional stereo method is used to generate proxy disparity labels, in which unreliable depth measurements are removed via a confidence measure to improve stereo accuracy. The disparity images are generated by training a dual encoder–decoder convolutional neural network from rectified stereo images coupled with proxy labels generated by the traditional stereo method. A principled mask is computed to exclude the pixels, which are not seen in one of views due to parallax effects from the calculation of loss function. Moreover, the neighbourhood smoothness term is employed to constrain neighbouring pixels with similar appearances to generate a smooth depth surface. This approach can make the depth of the projected point cloud closer to the real surgical site and preserve realistic details. The authors demonstrate the performance of the method by training and evaluation with a partial nephrectomy da Vinci surgery dataset and heart phantom data from the Hamlyn Centre.
Databáze: OpenAIRE
Externí odkaz: