-
The conjugated null space method of blind PSF estimation and deconvolution optimization
Authors:
Yuriy A. Bunyak,
Roman N. Kvetnyy,
Olga Yu. Sofina
Abstract:
We have shown that the vector of the point spread function (PSF) lexicographical presentation belongs to the left side conjugated null space (NS) of the autoregression (AR) matrix operator on condition the AR parameters are common for original and blurred images. The method of the PSF and inverse PSF (IPSF) evaluation in the basis of the NS eigenfunctions is offered. The optimization of the PSF an…
▽ More
We have shown that the vector of the point spread function (PSF) lexicographical presentation belongs to the left side conjugated null space (NS) of the autoregression (AR) matrix operator on condition the AR parameters are common for original and blurred images. The method of the PSF and inverse PSF (IPSF) evaluation in the basis of the NS eigenfunctions is offered. The optimization of the PSF and IPSF shape with the aim of fluctuation elimination is considered in NS spectral domain and image space domain. The function of surface area was used as the regularization functional. Two methods of original image estimate optimization were designed basing on maximum entropy generalization of sought and blurred images conditional probability density and regularization. The first method uses balanced variations of convolutions with the PSF and IPSF to obtaining iterative schema of image optimization. The variations balance is providing by dynamic regularization basing on condition of the iteration process convergence. The regularization has dynamic character because depends on current and previous image estimate variations. The second method implements the regularization of the deconvolution optimization in curved space with metric defined on image estimate surface. The given iterative schemas have fast convergence and therefore can be used for reconstruction of high resolution images series in real time. The NS can be used for design of denoising bilateral linear filter which does not introduce image smoothing.
△ Less
Submitted 26 February, 2015;
originally announced February 2015.
-
Blind PSF estimation and methods of deconvolution optimization
Authors:
Yu. A. Bunyak,
O. Yu. Sofina,
R. N. Kvetnyy
Abstract:
We have shown that the left side null space of the autoregression (AR) matrix operator is the lexicographical presentation of the point spread function (PSF) on condition the AR parameters are common for original and blurred images. The method of inverse PSF evaluation with regularization functional as the function of surface area is offered. The inverse PSF was used for primary image estimation.…
▽ More
We have shown that the left side null space of the autoregression (AR) matrix operator is the lexicographical presentation of the point spread function (PSF) on condition the AR parameters are common for original and blurred images. The method of inverse PSF evaluation with regularization functional as the function of surface area is offered. The inverse PSF was used for primary image estimation. Two methods of original image estimate optimization were designed basing on maximum entropy generalization of sought and blurred images conditional probability density and regularization. The first method uses balanced variations of convolution and deconvolution transforms to obtaining iterative schema of image optimization. The variations balance was defined by dynamic regularization basing on condition of iteration process convergence. The regularization has dynamic character because depends on current and previous image estimate variations. The second method implements the regularization of deconvolution optimization in curved space with metric defined on image estimate surface. It is basing on target functional invariance to fluctuations of optimal argument value. The given iterative schemas have faster convergence in comparison with known ones, so they can be used for reconstruction of high resolution images series in real time.
△ Less
Submitted 15 June, 2012;
originally announced June 2012.
-
Pattern recognition using inverse resonance filtration
Authors:
Olga Sofina,
Yuriy Bunyak,
Roman Kvetnyy
Abstract:
An approach to textures pattern recognition based on inverse resonance filtration (IRF) is considered. A set of principal resonance harmonics of textured image signal fluctuations eigen harmonic decomposition (EHD) is used for the IRF design. It was shown that EHD is invariant to textured image linear shift. The recognition of texture is made by transfer of its signal into unstructured signal whic…
▽ More
An approach to textures pattern recognition based on inverse resonance filtration (IRF) is considered. A set of principal resonance harmonics of textured image signal fluctuations eigen harmonic decomposition (EHD) is used for the IRF design. It was shown that EHD is invariant to textured image linear shift. The recognition of texture is made by transfer of its signal into unstructured signal which simple statistical parameters can be used for texture pattern recognition. Anomalous variations of this signal point on foreign objects. Two methods of 2D EHD parameters estimation are considered with the account of texture signal breaks presence. The first method is based on the linear symmetry model that is not sensitive to signal phase jumps. The condition of characteristic polynomial symmetry provides the model stationarity and periodicity. Second method is based on the eigenvalues problem of matrices pencil projection into principal vectors space of singular values decomposition (SVD) of 2D correlation matrix. Two methods of classification of retrieval from textured image foreign objects are offered.
△ Less
Submitted 16 March, 2010;
originally announced March 2010.
-
Location and Spectral Estimation of Weak Wave Packets on Noise Background
Authors:
Yu. Bunyak,
O. Bunyak
Abstract:
The method of location and spectral estimation of weak signals on a noise background is being considered. The method is based on the optimized on order and noise dispersion autoregressive model of a sought signal. A new approach of model order determination is being offered. Available estimation of the noise dispersion is close to the real one. The optimized model allows to define function of em…
▽ More
The method of location and spectral estimation of weak signals on a noise background is being considered. The method is based on the optimized on order and noise dispersion autoregressive model of a sought signal. A new approach of model order determination is being offered. Available estimation of the noise dispersion is close to the real one. The optimized model allows to define function of empirical data spectral and dynamic features changes. The analysis of the signal as dynamic invariant in respect of the linear shift transformation yields the function of model consistency. Use of these both functions enables to detect short-time and nonstationary wave packets at signal to noise ratio as from -20 dB and above.
△ Less
Submitted 2 July, 2007;
originally announced July 2007.
