Oleksiy Pogrebnyak

Digital adaptive robust algorithms for radar image filtering

Novel adaptive robust filtering algorithms applicable to radar image processing are proposed. They take into consideration the peculiarities of radar images and possess a good combination of properties: effective speckle suppression, impulsive noise removal, edge and detail preservation and low computational complexity. The advantages of these digital algorithms are demonstrated by simulated data and images obtained by airborne side-look non SAR radar.

Wiener discrete cosine transform-based image filtering

Abstract. A classical problem of additive white (spatially uncorrelated) Gaussian noise suppression in grayscale images is considered. The main attention is paid to discrete cosine transform (DCT)-based denoising, in particular, to image processing in blocks of a limited size. The efficiency of DCT-based image filtering with hard thresholding is studied for different sizes of overlapped blocks. A multiscale approach that aggregates the outputs of DCT filters having different overlapped block sizes is proposed. Later, a two-stage denoising procedure that presumes the use of the multiscale DCT-based filtering with hard thresholding at the first stage and a multiscale Wiener DCT-based filtering at the second stage is proposed and tested. The efficiency of the proposed multiscale DCT-based filtering is compared to the state-of-the-art block-matching and three-dimensional filter. Next, the potentially reachable multiscale filtering efficiency in terms of output mean square error (MSE) is studied. The obtained results are of the same order as those obtained by Chatterjee’s approach based on nonlocal patch processing. It is shown that the ideal Wiener DCT-based filter potential is usually higher when noise variance is high.

Image compression algorithm based on morphological associative memories

A new method for image compression based on Morphological Associative Memories (MAM) is proposed. We used MAM at the transformation stage of image coding, thereby replacing the traditional methods such as Discrete Cosine Transform or Wavelet Transform. After applying the MAM, the informative image data are concentrated in a minimum of values. The next stages of image coding can be obtained by taking advantage of this new representation of the image. The main advantage offered by the MAM with respect to the traditional methods is the speed of processing, whereas the compression rate and the obtained signal to noise ratios compete with the traditional methods.

Efficiency of DCT-Based Denoising Techniques Applied to Texture Images

Textures or high-detailed structures contain information that can be exploited in pattern recognition and classification. If an acquired image is noisy, noise removal becomes an operation to improve image quality before further stages of processing. Among possible variants of denoising, we consider filters based on orthogonal transforms, in particular, on discrete cosine transform (DCT) known to be able to effectively remove additive white Gaussian noise (AWGN). Besides, we study a representative of nonlocal denoising techniques, namely, BM3D known as state-of-the-art technique based on DCT and similar patch search. We show that noise removal in texture images using the considered DCT-based techniques can distort fine texture details. To detect such situations and avoid texture degradation due to filtering, we propose to apply filtering efficiency prediction tests applicable to wide class of images. These tests are based on DCT coefficient statistic parameters and can be used for decision-making in relation to the use of the considered filters.

Efficiency of texture image enhancement by DCT-based filtering

Textures or high-detailed structures as well as image object shapes contain information that is widely exploited in pattern recognition and image classification. Noise can deteriorate these features and has to be removed. In this paper, we consider the influence of textural properties on efficiency of image enhancement by noise suppression for the posterior treatment. Among possible variants of denoising, filters based on discrete cosine transform known to be effective in removing additive white Gaussian noise are considered. It is shown that noise removal in texture images using the considered techniques can distort fine texture details. To detect such situations and to avoid texture degradation due to filtering, filtering efficiency predictors, including neural network based predictor, applicable to a wide class of images are proposed. These predictors use simple statistical parameters to estimate performance of the considered filters. Image enhancement is analysed in terms of both standard criteria and metrics of image visual quality for various scenarios of texture roughness and noise characteristics. The discrete cosine transform based filters are compared to several counterparts. Problems of noise removal in texture images are demonstrated for all of them. A special case of spatially correlated noise is considered as well. Potential efficiency of filtering is analysed for both studied noise models. It is shown that studied filters are close to the potential limits.

Noise pattern recognition of airplanes taking off: task for a monitoring system

This paper presents an original work for aircraft noise monitoring systems and it analyzes the airplanes noise signals and a method to identify them. The method uses processed spectral patterns and a neuronal network feed-forward, programmed by means of virtual instruments. The obtained results, very useful in portable systems, make possible to introduce redundancy to permanent monitoring systems. The noise level in a city has fluctuations between 50 dB (A) and 100 dB (A). It depends on the population density and its activity, commerce and services in the public thoroughfare, terrestrial and aerial urban traffic, of the typical activities of labor facilities and used machinery, which give varied conditions that must be faced of diverse ways within the corresponding normalization. The sounds or noises that exceed the permissible limits, whichever the activities or causes that originate them, are considered events susceptible to degrade the environment and the health.

Approaches to classification of multichannel images

The comparison of different approaches to classification of multichannel remote sensing images obtained by spaceborne imaging systems is presented. It is demonstrated that it is reasonable to compress original noisy images with appropriate compression ratio and then to classify the decompressed images rather than original data. Two classifiers are considered: based on radial basis function neural network and support vector machine. The latter one produces slightly better classification results.

Wavelet transforms and neural networks applied to image retrieval

We face the problem of retrieving images from a database. During training a wavelet-based description of each image is first obtained using a Daubechies 4-wavelet transformation. Resulting coefficients are used to train a neural network (NN). During retrieval, a given image is presented to the already trained NN. The system responds with the most similar images. Three different ways to obtain the coefficients of the wavelet transform are tested: from the entire image, from the histogram of the biggest circular window inside the image color channels, and from the histograms of square sub-images in the image channels of the original image. 120 color images of airplanes were used for training and 240 for testing. The best efficiency of 88% was obtained with the third description method

Analysis of classification accuracy for pre-filtered multichannel remote sensing data

Noise is one of the main factors degrading the quality of original multichannel remote sensing data and its presence influences classification efficiency, object detection, etc. Thus, pre-filtering is often used to remove noise and improve the solving of final tasks of multichannel remote sensing. Recent studies indicate that a classical model of additive noise is not adequate enough for images formed by modern multichannel sensors operating in visible and infrared bands. However, this fact is often ignored by researchers designing noise removal methods and algorithms. Because of this, we focus on the classification of multichannel remote sensing images in the case of signal-dependent noise present in component images. Three approaches to filtering of multichannel images for the considered noise model are analysed, all based on discrete cosine transform in blocks. The study is carried out not only in terms of conventional efficiency metrics used in filtering (MSE) but also in terms of multichannel data classification accuracy (probability of correct classification, confusion matrix). The proposed classification system combines the pre-processing stage where a DCT-based filter processes the blocks of the multichannel remote sensing image and the classification stage. Two modern classifiers are employed, radial basis function neural network and support vector machines. Simulations are carried out for three-channel image of Landsat TM sensor. Different cases of learning are considered: using noise-free samples of the test multichannel image, the noisy multichannel image and the pre-filtered one. It is shown that the use of the pre-filtered image for training produces better classification in comparison to the case of learning for the noisy image. It is demonstrated that the best results for both groups of quantitative criteria are provided if a proposed 3D discrete cosine transform filter equipped by variance stabilizing transform is applied. The classification results obtained for data pre-filtered in different ways are in agreement for both considered classifiers. Comparison of classifier performance is carried out as well. The radial basis neural network classifier is less sensitive to noise in original images, but after pre-filtering the performance of both classifiers is approximately the same.

A Fast Search Algorithm for Vector Quantization Based on Associative Memories

One of the most serious problems in vector quantization is the high computational complexity at the encoding phase. This paper presents a new fast search algorithm for vector quantization based on Extended Associative Memories (FSA-EAM). In order to obtain the FSA-EAM, first, we used the Extended Associative Memories (EAM) to create an EAM-codebook applying the EAM training stage to the codebook produced by the LBG algorithm. The result of this stage is an associative network whose goal is to establish a relation between training set and the codebook generated by the LBG algorithm. This associative network is EAM-codebook which is used by the FSA-EAM. The FSA-EAMVQ process is performed using the recalling stage of EAM. This process generates a set of the class indices to which each input vector belongs. With respect to the LBG algorithm, the main advantage offered by the proposed algorithm is high processing speed and low demand of resources (system memory), while the encoding quality remains competitive.