Unknown Pixels Research Articles

Designing of a type-2 fuzzy logic filter for improving edge-preserving restoration of interlaced-to-progressive conversion

This paper focuses on recently advanced fuzzy models and the application of type-2 fuzzy sets in video deinterlacing. The final goal of the proposed deinterlacing algorithm is to exactly determine an unknown pixel value while preserving the edges and details of the image. To begin, we will discuss some artefacts of spatial, temporal, and spatio-temporal domain deinterlacing methods. In order to address the aforementioned issues, we adopted type-2 fuzzy sets concepts to design a weight evaluating approach. In the proposed method, the upper and lower fuzzy membership functions of the type-2 fuzzy logic filters are derived from the type-1 (or primary) fuzzy membership function. The weights from upper and lower membership functions are considered to be multiplied with the candidate deinterlaced pixels. Experimental results proved that the performance of the proposed method was superior, both objectively and subjectively to other different conventional deinterlacing methods. Moreover, the proposed method preserved the smoothness of the original image edges and produced a high-quality progressive image.

Robust reconstruction of low-resolution document images by exploiting repetitive character behaviour

In this paper, we present a new approach for reconstructing low-resolution document images. Unlike other conventional reconstruction methods, the unknown pixel values are not estimated based on their local surrounding neighbourhood, but on the whole image. In particular, we exploit the multiple occurrence of characters in the scanned document. In order to take advantage of this repetitive behaviour, we divide the image into character segments and match similar character segments to filter relevant information before the reconstruction. A great advantage of our proposed approach over conventional approaches is that we have more information at our disposal, which leads to a better reconstruction of the high-resolution (HR) image. Experimental results confirm the effectiveness of our proposed method, which is expressed in a better optical character recognition (OCR) accuracy and visual superiority to other traditional interpolation and restoration methods.

Hyperspectral target detection using Gaussian filter and post-processing

A novel target detection technique is developed in this paper for hyperspectral imagery. The proposed technique involves a Gaussian filter to process the reference target signature as well as the unknown pixel signature from the given input scene. After performing correlation operation between the signatures, a post-processing scheme is employed to decide whether the specific pixel contains the target material. Theoretical analyses and computer simulation results presented in this paper show that the technique can successfully overcome the problem of variations in the target signature due to changes in the environment and addition of noise to the input scene. Investigation on real-life hyperspectral imagery proves that the proposed technique can successfully detect the potential targets present in the input scene without generating any significant amount of false alarm.

A Geostatistically Based Preprocessing Algorithm for Hyperspectral Data Analysis

We present a preprocessing algorithm for hyperspectral remote sensing datasets. The algorithm is based on a geostatistical method and should be helpful when a spatial relationship is detected in a dataset. One significant advantage of hyperspectral remote sensing using spectral profiles is the ability to compare an unknown pixel's profile with endmembers that have already been identified by a variety of methods (e.g., laboratory experiments with high-precision spectrometers), with the final goal of determining the unknown pixel. The conditions under which the airborne or spaceborne hyperspectral data are collected, however, differ from those that prevail in the laboratory or field. Therefore, a dataset should be preprocessed so as to eliminate or considerably reduce these differences; the algorithm presented here could be used for that purpose. The result will not only improve the smoothness of spectral profiles, but it may also offer advantages for geological investigations to study mineral anomalies using hyperspectral data. Concentrations of minerals in rock bodies often have certain patterns and follow trends that can be modeled by computing a semivariogram. The advantages of using such a trend have induced mining engineers to develop innovations in geostatistics. These trends should be taken into account when handling hyperspectral datasets. In all methods presented for boosting spectral profiles, the spatial relationships among pixels' DNs are neglected, but, in the method presented here, this relationship is calculated by geostatistics, and an algorithm is applied to improve spectral profiles. The nugget effect is calculated separately for each channel, and its square root is subtracted from the reflectance of all pixels in that channel. Finally, we examine the effectiveness and validation of the method examined using the AVIRIS dataset from Cuprite, Nevada. The results are satisfactory, as the algorithm yields a better mineral detection process.

Local spatial properties based image interpolation scheme using SVMs

Image interpolation plays an important role in image process applications. A novel support vector machines (SVMs) based interpolation scheme is proposed with increasing the local spatial properties in the source image as SVMs input patterns. After the proper neighbor pixels region is selected, trained support vectors are obtained by training SVMs with local spatial properties that include the average of the neighbor pixels gray values and the gray value variations between neighbor pixels in the selected region. The support vector regression machines are employed to estimate the gray values of unknown pixels with the neighbor pixels and local spatial properties information. Some interpolation experiments show that the proposed scheme is superior to the linear, cubic, neural network and other SVMs based interpolation approaches.

Fuzzy Weighted Approach to Improve Visual Quality of Edge-Based Filtering

A new algorithm for the deinterlacing of interlaced scanned video sequences is presented in this paper. The ultimate goal of the proposed algorithm is to exactly determine the unknown pixel value while preserving the edges and details of the image. Traditional edge direction-based interpolation algorithms usually introduce artifacts in the case of non-clear edges. In order to solve the above issues, we adopted fuzzy concepts to design a weight-evaluating technique. The weights were considered to be multiplied by the candidate deinterlaced pixels. Experimental results demonstrated that the resulting frames contained less errors and sharper edges than frames generated using common approaches such as median filtering and edge direction-based linear interpolation.

A Markov Random Field Model-Based Approach to Natural Image Matting

This paper proposes a Markov Random Field (MRF) model-based approach to natural image matting with complex scenes. After the trimap for matting is given manually, the unknown region is roughly segmented into several joint sub-regions. In each sub-region, we partition the colors of neighboring background or foreground pixels into several clusters in RGB color space and assign matting label to each unknown pixel. All the labels are modelled as an MRF and the matting problem is then formulated as a maximum a posteriori (MAP) estimation problem. Simulated annealing is used to find the optimal MAP estimation. The better results can be obtained under the same user-interactions when images are complex. Results of natural image matting experiments performed on complex images using this approach are shown and compared in this paper.

Multigrid tomographic inversion with variable resolution data and image spaces

A multigrid inversion approach that uses variable resolutions of both the data space and the image space is proposed. Since the computational complexity of inverse problems typically increases with a larger number of unknown image pixels and a larger number of measurements, the proposed algorithm further reduces the computation relative to conventional multigrid approaches, which change only the image space resolution at coarse scales. The advantage is particularly important for data-rich applications, where data resolutions may differ for different scales. Applications of the approach to Bayesian reconstruction algorithms in transmission and emission tomography with a generalized Gaussian Markov random field image prior are presented, both with a Poisson noise model and with a quadratic data term. Simulation results indicate that the proposed multigrid approach results in significant improvement in convergence speed compared to the fixed-grid iterative coordinate descent method and a multigrid method with fixed-data resolution.

Edge-directed interpolation based on shifted-linear image model

This paper presents a hybrid image interpolation algorithm to keep details and edges simultaneously. The basic idea is to separate the unknown pixels into two classes and estimate them in different way. One class of unknown pixels is obtained via shifted linear interpolation and the other class through statistical signal processing method. The merit of this hybrid algorithm is that each unknown pixel can be estimated through original pixels simultaneously. Simulation results demonstrate that this hybrid interpolation algorithm improves the quality of the interpolated images over conventional interpolation methods.

A Weighted Linear Extrapolation-Based Simple Image Enlargement Algorithm

Abstract In this paper we propose a very fast and high quality image enlazgement algorithm, which estimates unknown pixel values in the image by using a weighted sum of linear extrapolations of the pixels neighboring to the pixel to be estimated, and reconstructs the shazp edges, peaks and monotonic slopes on the enlarged image. The validity and the effectiveness of the proposed algorithm aze verified by applying it to the enlargement of the actual digital images. The experimental results show that the performance of the proposed algorithm is superior to other typical image enlazgement algorithms objectively and subjectively and, in addition, the processing time for enlarging image is reduced drastically in comparison with other algorithms which achieve the image enlargement with almost same quality to our algorithm previously proposed [1].

An image zooming technique based on vector quantization approximation

An image zooming method based on vector quantization approximation for magnifying gray-scale and color image by a factor of 2 is proposed. In our proposed method, the unknown pixel values on the image are interpolated by using a vector quantization codebook based on their local information. In comparison of our method with the locally adaptive zooming algorithm published in [S. Battiato, G. Gallo, F. Stanco, A locally adaptive zooming algorithm for digital images, Image and Vision Computing, 20(11) (2002) 805–812.], our experimental results have demonstrated that the image quality of the enlarged image is superior to the method in [S. Battiato, G. Gallo, F. Stanco, A locally adaptive zooming algorithm for digital images, Image and Vision Computing, 20(11) (2002) 805–812.]. Not only is our method simpler to implement by utilizing a table look-up technique on codebook, but also is much easier in translating to color images than that of [S. Battiato, G. Gallo, F. Stanco, A locally adaptive zooming algorithm for digital images, Image and Vision Computing, 20(11) (2002) 805–812.] by replacing an adequate codebook.

Bayesian inversion of imaging spectrometer data using a fuzzy geological outcrop model

In geological remote sensing, in particular when studying the Earth with use of high spectral resolution sensors (so-called imaging spectrometers), ‘mineral maps' are the outcome of the image processing. Such mineral maps portray the spatial arrangement of classes of surface mineralogy derived from a comparison between known mineral signatures from spectral libraries and unknown (imaged in the hyperspectral domain) pixel spectra. With this information, geologists can build up a geological model of an area. Here, the reverse approach is advocated by introducing the geological model (where geology is converted to physical measurements by introducing spectra) at the start of the hyperspectral processing chain. This model is refined in an iterative way by comparing its mismatch with the observed reflectance measurements. Hence, a (geophysical) inversion of the hyperspectral dataset to a geological model is done. In a previous article explaining the physics and mathematics of the Bayesian inversion techniques, a ‘rigid geological model’ was used. Although the model takes on the advantage of including spatial arrangement and contextual information, hard field evidence is not included. Here an extension to this model is presented in the form of a ‘fuzzy outcrop geological model’. It is shown that the fuzzy outcrop model, which uses known outcrops and inferred outcrop patterns, outperforms the rigid geological model.

Two nonlinear inverse methods for electromagnetic induction measurements

The authors develop two nonlinear inverse methods to reconstruct the conductivity profile from electromagnetic induction (EMI) measurements: the improved two-step inverse method based on the extended Born approximation (EBA) and the combination of the EBA and the contrast source inversion (CSI) method. In the first method, the nonlinear problem is recast as a two-step linear inversion and is solved by using the extended Born approximation. The authors improve this method with the fast Fourier transform (FFT) algorithm and by a conjugate-gradient optimization procedure. For a problem with N unknown pixels and ill measurement data points, the computational time of this inverse EBA procedure is reduced to O(MN) in the first step (i.e., the underdetermined linear problem) and to O(Nlog/sub 2/N) in the second step (i.e., the well-determined linear problem) using the newly developed FFT-EBA method. Furthermore, the memory requirement is reduced to O(M/sub N/) (M/sub T/ is the number of transmitters). In the second inverse method, they apply the contrast source inversion (CST) method, but use the two-step linear inversion result as the initial solution. By using FFT, the CPU time costs O(C/sub 1/MN+C/sub 2/N log/sub 2/ N) in each iteration of the CSI procedure. Numerical examples illustrate the effectiveness of these methods, even up to a high contrast of 100:1.

Geophysical inversion of imaging spectrometer data for geologic modelling

Imaging spectrometers acquire physical measurements of radiation in many small and contiguous wavebands in order to produce continuous spectral information that can be compared directly with measurements made in the field or laboratory if the atmospheric attenuation is compensated for. Widely used processing strategies compare unknown pixel reflectance spectra with known field or laboratory spectra and derive statistical measures of the goodness-of-fit of these measurements. In geological remote sensing practice these yield 'mineral maps' that on a pixel-by-pixel basis portray the similarity of the measured signal with that of signatures of known minerals. With this information, geologists can build up a geological model of an area. Here the reverse processing strategy is proposed, introducing the geologic model (where geology is converted to physical measurements by introducing spectra) at the start of the processing chain and refining the model in an iterative way by comparing its mismatch with observed reflectance measurements.

Validated surface mineralogy from high‐spectral resolution remote sensing: a review and a novel approach applied to gold exploration using AVIRIS data

Remote sensing’s main limitation to date was that geological surface information lacked detail due to the broad bandwidth of the operational scanners. Imaging spectrometers now allow acquisition of image data in many contiguous spectral bands, with the aim of producing laboratory‐like reflectance spectra for each pixel in the scene which can be directly compared with spectra of known materials measured in the field or laboratory to allow mineral identification on the Earth surface. The advent of satellite‐based imaging spectrometer systems justifies the surge for techniques that allow the automatic preparation of validated surface mineralogy maps. We show an extension of a procedure known as cross‐correlogram spectral matching to retrieve accuracy assessed surface mineralogy from Airborne Visible/InfraRed Imaging Spectrometer data (AVIRIS). The procedure is based on the spectral cross‐correlation of known mineral spectra with unknown pixel spectra from AVIRIS. The validation is performed through a root‐mean square calculation of expected correlograms and pixel correlograms. Results obtained in the study outlined allow an accurate delineation of zones of hydrothermal alteration in an area of active gold exploration.

Point accuracy of a non-parametric method in estimation of forest characteristics with different satellite materials

Abstract The reference sample plot (RSP) method is a distance-weighted k nearest neighbour estimation method, which allows simultaneous interpretation of several variables. In the RSP method, the k spectrally nearest field plots are looked at separately for each unknown pixel, and the area weight of the unknown pixel is divided as a function of the spectral distances to the nearest plots. The RSP method was examined in a forest inventory for estimating stem volumes by tree species groups using different satellite materials. Two methods were tested both in searching for and weighting the nearest field plots. Euclidean distance functions worked steadily with all the volume variables studied. The other distance measure tested was based on regression modelling. With more than 15 plots, both covariance weighting and inverse distance weighting gave similar results. Considering the field data of this study, the suitable number of the nearest plots in plotwise estimation appeared to be between 10 and 15 plots. Wi...

Classification of remotely-sensed imagery using an indicator kriging approach: application to the problem of calcite-dolomite mineral mapping

Abstract A new image classification technique for analysis of remotely-sensed data based on geostatistical indicator kriging is introduced. Conventional classification techniques require ground truth information, use only the spectral characteristics of an unknown pixel in comparison, rely on a Gaussian probability distribution for the spectral signature of the training data, and work on a pixel support or spatial resolution without allowing classification on larger or smaller volumes. The indicator kriging classifier overcomes such problems because: (1) it relies on spectral information from laboratory studies rather than on ground truth data, (2) through the kriging estimation variances an estimate of uncertainly is derived, (3) it incorporates spatial aspects because it uses local estimation techniques, (4) it is distribution-free, (5) and may be applied on different supports if the data are corrected for support changes. Comparison of classification results applied to the problem of mapping calcite and dolomite from GER imaging spectrometry data shows that indicator kriging performs better than the conventional classification algorithms and gives insight in the accuracy of the results without prior field knowledge

Performance characteristics of the indicator classifier on simulated image data

Classification of remotely sensed imagery into groups of pixels having similar spectral reflectance characteristics is conducted classically by comparing the spectral signature of unknown pixels with those of training pixels of known ground cover type. Thus classification methods use only the spectral characteristics of image data without considering the spatial aspects or the relative location of an unknown pixel with respect to pixels from the training data set. An indicator classifier was introduced in 1992 that combines spatial and spectral information in a decision model. In this Letter the performance of this classifier is tested on simulated image data with known mineral targets and varying spatial variability and noise. It is demonstrated that incorporating spatial continuity into the classification process may largely increase the accuracy of the resulting classified images.