Dong Yin

SAR Image Compression Using Multiscale Dictionary Learning and Sparse Representation

In this letter, we focus on a new compression scheme for synthetic aperture radar (SAR) amplitude images. The last decade has seen a growing interest in the study of dictionary learning and sparse representation, which have been proved to perform well on natural image compression. Because of the special techniques of radar imaging, SAR images have some distinct properties when compared with natural images that can affect the design of a compression method. First, we introduce SAR properties, sparse representation, and dictionary learning theories. Second, we propose a novel SAR image compression scheme by using multiscale dictionaries. The experimental results carried out on amplitude SAR images reveal that, when compared with JPEG, JPEG2000, and a single-scale dictionary-based compression scheme, the proposed method is better for preserving the important features of SAR images with a competitive compression performance.

Compression technique for compressed sensing hyperspectral images

The compressed sensing (CS) theorem is a novel sampling approach that breaks through the conventional Nyquist sampling limit and brings a revolution in the field of signal processing. This article investigates the compression technique for CS hyperspectral images so as to illustrate the superiority provided by this new theorem. First, several comparative experiments are used to reveal that the drawback of prior compression techniques, designed for the data acquired by the conventional hyperspectral imaging system, is either low compression ratio or a waste of sampling resource. After a condensed analysis, we state that the CS theorem provides the probability of avoiding such defects. Then a straightforward scheme, which takes advantage of spectral correlation, is proposed to compress the CS hyperspectral images to reduce the data size further. Moreover, a flexible recovery strategy is designed to speed up the reconstruction of original bands from the corresponding CS images. The experimental results based on the actual hyperspectral images have demonstrated the efficiency of this proposed technique.

Hyperspectral data compression using sparse representation

Due to all bands of hyperspectral data have the same imaging area, it is reasonable to believe that the dictionary can sparse represent one band may also represent the other bands sparsely. Based on this property, this paper presents a new compression frame for hyperspectral data using sparse representation, and a simplified algorithm under this frame is also proposed. The basic idea of the proposed algorithm is to sparse coding bands using the dictionary learned from one training band, and its innovation is that patches having the same spatial location of all bands are restricted to be represented using the same atoms. Experimental results based on OMP and K-SVD are provided, which reveal that this proposal has better performance than wavelet based compression algorithm at low bit rates.

An adaptive sparse representation for remote sensing image based on combination of wavelet and adaptive directional filter

This paper presents an adaptive sparse representation scheme for the remote sensing image, the geometric structure of which is more complex than that of natural image. The presented scheme includes two main stages which are wavelet transform and adaptive directional filter which is designed based on a binary tree. The construction of the binary tree depends on the image geometric information in frequency domain. Besides the properties of multiscale, multidirectionality and nonredundancy, our scheme has an additional property named adaptation which is a particular characteristic. Experimental results show that, in the sense of signal to noise ratio and visual quality, the performance of the nonlinear approximation of our scheme for remote sensing image is better than most of the existing sparse representation schemes.

Bridge recognition of median-resolution SAR images using pun histogram entropy

A novel algorithm for bridge recognition of median synthetic aperture radar (SAR) images using histogram entropy presented by Pun is proposed. Firstly, Lee filter and histogram proportion are used to denoise the original image and to make the target evident. Then, water regions are gained through histogram segmentation and the contours of water regions are extracted. After these, the potential bridge targets are obtained based on the space relativity between bridges and water regions using improved contour search. At last, bridges are recognized by extracting the feature of Pun histogram entropy (PHE) of these potential bridge targets. Experimental results show the good qualities of the algorithm, such as fast speed, high rate of recognition, and low rate of false target.

Medical Image Categorization Based on Gaussian Mixture Model

In this paper we present an approach for medical image categorization based on Gaussian mixture model. There are distinct differences on texture, shape and intensity characteristics among the images of different parts of body. Considering of the features of the Gaussian mixture model , first we extract the characteristic vectors of the training image set to learn the class model for each class , then categorize the test image using the Bayesian principle. The experimental results indicate that the method performs very well on CT image categorization. We achieved classification accuracy up to 97% in the experiment.

Target recognition application research with gradient magnitude and direction

The paper presents a target recognition method with magnitude and direction of gradient in mid-resolution optical remote sensing images. Firstly, the directional texture is analyzed and a fast location method based on the gradient blurring process is realized. Secondly, the target characteristic like geography, gray, geometry are analyzed and described. Finally, using the shape character distance, the target is successfully matched, measured and recognized. Experiment results show that this automatic recognition method is effective, and the new location algorithm of the common directional texture based on gradient blurring is very promising and potentially useful for the sequent recognitions of targets.

Traffic objects auto-recognition research based-on MAS technology

How to find an efficient method to solve the problems of object auto-recognition has been a research hot point. For the application of remote sensing image processing, this paper analyses features and inner relations of traffic objects. Handling with multi-agent system (MAS) technology, by using prior knowledge, we design a whole auto-recognition pattern. That is based-on the thought of dynamic plan to determine the road seed, and extract the road information. With it, we can extract the bridge, toll-station and tunnel mouth. Conversely, based-on the thought of separation ocean from land, we can extract the bridge kindly, and then track the road. So, a method of intelligent recognition with mutual inference is formed. For the size more than 10000times10000 pixels remote sensing images, the recognition results of road, bridge, toll-station and tunnel are good performance.

Categorization Method Research for Medical Image Using Gaussian Mixture Model

The paper presents an approach for medical image categorization based-on Gaussian mixture model in CBMIR system. The medical image categorization is a very complicated problem because the characteristics on texture, shape and intensity among the images of different parts of body are distinct differences. First, we extract the characteristic vectors of the training image set. Then, we choose the optimum features which can distinguish different classes and the same class better. After getting GMM parameters by EM algorithm, we categorize the test images. The experimental results indicate that the method performs well on CT image categorization.

Multi-scale feature analysis method for bridge recognition in SAR images

Target recognition is an important field in remote sensing image processing. This paper handles with the problem of bridge recognition in Synthetic Aperture Radar (SAR) images. Based on features of bridges, rivers and land in different spatial resolution SAR images, a method of multi-scale analysis is proposed. The original data is split into low and middle resolution level. The bridge candidates are located in low level by automatically detecting river. And regions of interesting (ROI) are obtained in middle one. So, bridges are recognized by analyzing those regions. The example results indicate that the processing speed can be greatly improved and the precision of recognition can also be ensured.