Showing papers by "Rishikesh Kulkarni published in 2022"

Objective speckle pattern-based surface roughness measurement using matrix factorization.

Abstract: A method for the measurement of profile parameters of both isotropic and anisotropic surfaces is presented using the objective laser speckle imaging technique. The surface parameters are characterized in terms of a singular value decomposition method-based metric derived from the initial key contributing singular values of the speckle pattern. A simulation study is performed with random Gaussian anisotropic surfaces generated as a function of the correlation lengths in both x and y directions. In the experimental demonstration, the proposed method is verified with metallic samples having distinct surface roughness processed through widely used machining operations viz., vertical milling, and grinding. A brief discussion about the extent to which the minimum number of singular values that are sufficient to evaluate the profile parameters in the context of experimental results is provided. The method supports the measurement of profile parameters of higher magnitude in the realm of non-contact topographic measurement techniques. The experimental results substantiate the practical applicability of the proposed method.

Phase unwrapping algorithm using breadth-first-search and multi-level segmentation of phase quality interval in digital holography

Abstract: A quality-guided phase unwrapping algorithm is proposed based on the breadth-first-search and multi-level segmentation of the phase quality interval. In the proposed method, the pixels having phase quality values below a threshold associated with a given segment are unwrapped based on the breadth-first-search strategy implemented in a recursive manner. The pixel selection based on the multi-level phase quality interval segmentation allows to perform noise-robust phase unwrapping and the recursive breadth-first-search technique offers computational efficiency in the implementation of proposed algorithm. Three segmentation strategies are investigated in reference to the phase unwrapping accuracy. Simulation and experimental results indicate that the proposed algorithm offers desirable trade-off between the phase unwrapping accuracy and computation time.

Weakly Supervised Learning-based Reconstruction of Planktons in Digital In-line Holography

Abstract: Weakly-supervised learning-based technique is employed for hologram reconstruction of planktons recorded in digital in-line holographic setup. One-level wavelet decomposition of the reconstructed image is computed to generate a mask and subsequently a ground truth image to train the network.

Fringe pattern demodulation using Zernike polynomials and a l1-norm regularized extended Kalman filter.

Abstract: A novel algorithm for closed fringe demodulation for an absolute phase estimation, to the best of our knowledge, is proposed. The two-dimensional phase is represented as a weighted linear combination of a certain number of Zernike polynomials (ZPs). Essentially, the problem of phase estimation is converted into the estimation of ZP coefficients. The task of ZP coefficient estimation is performed based on a state space model. Due to the nonlinear dependence of the fringe intensity measurement model on the ZP coefficients, the extended Kalman filter (EKF) is used for the state estimation. A pseudo-measurement model is considered based on the state vector sparsity constraint to improve the convergence performance of the EKF. Simulation and experimental results are provided to demonstrate the noise robustness and the practical applicability of the proposed method.

Wrapped phase denoising using adaptive Kalman smoother algorithm

Abstract: A wrapped phase pattern denoising algorithm is proposed based on the adaptive Kalman filtering. The exponential phase field (EPF) corresponding to the noisy wrapped phase pattern is considered as the measurement data. Row-wise and column-wise phase denoising is performed in a sequential manner. In each row/column, the EPF is denoised using the Kalman filter considering it as a spatially varying auto-regressive process. For achieving measurement independent, automatic phase denoising, the system and measurement noise covariances are adaptively estimated in the implementation of Kalman filter. The proposed formulation also allows simultaneous denoising of multiple rows/columns to reduce the computation cost. The amplitude of denoised EPF can be used as a phase quality metric which can aid in the subsequent phase unwrapping operation. The denoising performance is quantitatively evaluated using simulation results. Experimental results are provided to further substantiate the practical applicability of the proposed method.