China 
Africa 
Explore chapters and articles related to this topic
2+ Imaging
Published in Francesco S. Pavone, Shy Shoham, Handbook of Neurophotonics, 2020
Tobias Nöbauer, Alipasha Vaziri
Next, SID reconstructs the standard deviation image, or the individual NMF components by Richardson–Lucy deconvolution with the simulated LFM PSF, as described in Section 4.2.3. Depending on the data, it may be helpful to add sparsity and/or smoothness constraints – such as the well-known total variation regularization (Dey et al., 2004) – to the reconstruction problem to achieve more robust and artefact-free results. The reconstructed volumes are then segmented by performing a local maximum search. This yields an initial guess of neuron locations. Each neuron location estimate is used to generate a template of camera pixels that may receive scattered contributions from this neuron. To do so, a 3D Gaussian brightness distribution is placed centered on each individual neuron candidate location and projected forward by convolution with the LFM PSF. The resulting simulated LFM camera image is thresholded, resulting in a template for each neuron.
Fast Dual Optimization for Medical Image Segmentation
Published in Ayman El-Baz, Jasjit S. Suri, Big Data in Multimodal Medical Imaging, 2019
Jing Yuan, Ismail Ben Ayed, Aaron Fenster
During the last 20 years, convex optimization was successfully introduced as a powerful tool in image processing, computer vision and machine learning, which is mainly credited to the pioneering works from both theoretical and algorithmic studies [1–7], along with vast applications [8–13]. Typical applications include, for example, the total-variation-based image denoising [8,12] minu∫D(u−f)dx+α∫|∇u|dx,
Registration for Super-Resolution: Theory, Algorithms, and Applications in Image and Mobile Video Enhancement
Published in Peyman Milanfar, Super-Resolution Imaging, 2017
Patrick Vandewalle, Luciano Sbaiz, Martin Vetterli
where β is a regularization term that makes T differentiable (we used β = 0.01 for frames with values in the range [0 , 255]). The value of the Total Variation is related to the gradient magnitude of the image intensity, which is a measure of the image sharpness. In this way, the second term of equation (6.28) limits the amount of high frequencies added by the algorithm. In a software implementation of the Total Variation, the gradient operator in equation (6.29) is replaced by a difference operator. In our experiments, we approximated the derivatives with the average of the forward and backward differences along the x and the y coordinates. The constant α in (6.28) controls the trade-off between regularity and level of details in the output image.
An advanced fuzzy C-Means algorithm for the tissue segmentation from brain magnetic resonance images in the presence of noise and intensity inhomogeneity
Published in The Imaging Science Journal, 2023
Sandhya Gudise, K. Giri Babu, T. Satya Savithri
The noise effect can be decreased using spatial information while defining the cost function. The performance of the modified BCFCM can be further increased using an effective denoising algorithm that respects the edges. The total Variation (TV) denoising technique is employed in this proposed work to reduce the noise and protect the edges. TV denoising is applied to the membership function of the image in each iteration. Denoising helps in a better approximation of which helps in computing better values for and . During the next iteration, better values of and give the best values for and this process will be continued which results in fast convergence. The Chaotic-firefly algorithm is used in selecting better initial cluster representatives and then the results are used as the initial cluster representatives for the next step of the modified BCFCM technique and help in achieving good results and convergence. As mentioned in the introduction FA algorithm is combined with FCM to overcome the disadvantages of FCM.
A hyperautomative human behaviour recognition algorithm based on improved residual network
Published in Enterprise Information Systems, 2023
Jianxin Li, Jie Liu, Chao Li, Fei Jiang, Jinyu Huang, Shanshan Ji, Yang Liu
By introducing stream information, video frames, sparse sampling and non-local convolution into 3D-ResNet50 network, this paper proposes a new NLFT-3DResNet network model. NLFT-3DResNet is improved on the basis of 3D-ResNet50. By adding time dimension, it can better express the time sequence of features. Total Variation-L1 model is used to extract optical flow information, integrate nonlocal information, obtain inter-frame information, and extract the relevant attributes between features. The interference during network training is reduced by sparse sampling with non-fixed step size method. It can remove the residual information between adjacent frames and appropriately increase the number of layers of the non-local convolution network to ensure accuracy, better characterise the space-time characteristics, and improve the recognition rate.
Reconstruction algorithm for 3D Compton scattering imaging with incomplete data
Published in Inverse Problems in Science and Engineering, 2021
In the next section, we will see that the proposed OSEM algorithm provides satisfactory results and remains stable with Poisson noise. However, the algorithm does not denoise the reconstructed image. For this purpose, we add at the end of the algorithm a denoising step based on total variation. A standard method, see [44], consists in finding a minimizer of the functional with the noisy 3D image, λ a regularization parameter and Ω the compact domain of the 3D object under study. The construction of the minimizer can be expressed as the following gradient descent flow: where the parameter ε apodizes the singularity when . For the simulation results, we computed 1000 steps with , and .