[The meteorological satellite spectral image registration based on Fourier-Mellin transform].

The meteorological satellite spectral image is an effective tool for researches on meteorological science and environmental remote sensing science. Image registration is the basis for the application of the meteorological satellite spectral image data. In order to realize the registration of the satellite image and the template image, a new registration method based on the Fourier-Mellin transform is presented in this paper. Firstly, we use the global coastline vector map data to build a landmark template, which is a reference for the meteorological satellite spectral image registration. Secondly, we choose infrared sub-image of no cloud according to the cloud channel data, and extract the edges of the infrared image by Sobel operator. Finally, the affine transform model parameters between the landmark template and the satellite image are determined by the Fourier-Mellin transform, and thus the registration is realized. The proposed method is based on the curve matching in essence. It needs no feature point extraction, and can greatly simplify the process of registration. The experimental results using the infrared spectral data of the FY-2D meteorological satellite show that the method is robust and can reach a high speed and high accuracy.

[A coarse-to-fine registration method for satellite infrared image and visual image].

In the present paper, in order to resolve the registration of the multi-mode satellite images with different signal properties and features, a two-phase coarse-to-fine registration method is presented and is applied to the registration of satellite infrared images and visual images. In the coarse registration phase of this method, the edge of infrared and visual images is firstly detected. Then the Fourier-Mellin transform is adopted to process the edge images. Finally, the affine transformation parameters of the registration are computed rapidly by the transformation relation between the registering images in frequency domain. In the fine registration phase of the proposed method, the feature points of infrared and visual images are firstly detected by Harris operator. Then the matched feature points of infrared and visual images are determined by the cross-correlation similarity of their local neighborhoods. The fine registration is finally realized according to the spatial correspondent relation of the matched feature points in infrared and visual images. The proposed coarse-to-fine registration method derives both the advantages of two methods, the high efficiency of Fourier-Mellin transform based registration method and the accuracy of Harris operator based registration method, which is considered the novelty and merit of the proposed method. To evaluate the performance of the proposed registration method, the coarse-to-fine registration method is implemented on the infrared and visual images captured by the FY-2D meteorological satellite. The experimental results show that the presented registration method is robust and has acceptable registration accuracy.

[Spectral classification based on Bayes decision].

The rapid development of astronomical observation has led to many large sky surveys such as SDSS (Sloan digital sky survey) and LAMOST (large sky area multi-object spectroscopic telescope). Since these surveys have produced very large numbers of spectra, automated spectral analysis becomes desirable and necessary. The present paper studies the spectral classification method based on Bayes decision theory, which divides spectra into three types: star, galaxy and quasar. Firstly, principal component analysis (PCA) is used in feature extraction, and spectra are projected into the 3D PCA feature space; secondly, the class conditional probability density functions are estimated using the non-parametric density estimation technique, Parzen window approach; finally, the minimum error Bayes decision rule is used for classification. In Parzen window approach, the kernel width affects the density estimation, and then affects the classification effect. Extensive experiments have been performed to analyze the relationship between the kernel widths and the correct classification rates. The authors found that the correct rate increases with the kernel width being close to some threshold, while it decreases with the kernel width being less than this threshold.

[Redshift estimation of galaxy spectra based on similarity measure].

Automated spectra analysis is desirable and necessary for efficiency of large sky surveys such as SDSS (Sloan digital sky survey), 2DF (2 degree fields) and LAMOST (large sky area multi-object spectroscopic telescope). In the present paper, we present a method for redshift estimation of galaxy spectra based on similarity measure. Firstly, we extract the spectral lines of the observed spectrum using the feature constrains of spectral lines; secondly, the authors determine the redshift candidates of the observed spectrum by spectral line features; then, the similarity between the observed spectrum and the template spectra shifted by each redshift candidate is measured; finally, the candidate of the highest similarity is chosen as the estimated redshift. PCA (principal component analysis) is used to build the static galaxy template spectra. The authors perform PCA for the four template spectra E, S0, Sa and Sb of the normal galaxy and the seven template spectra Sc, Sb1, Sb2, Sb3, Sb4, Sb5 and Sb6 of the starburst galaxy respectively, where the eleven template spectra are presented by Kinney & Calzetti et al. Two eigen-spectra are produced with the variance contribution rate of 99%. The authors choose the two eigen-spectra as the galaxy templates. The similarity measure proposed, which is similar to the evidence accumulation, is defined as the weighted sum of several similarity evidences. It can reduce the influence caused by some error matching. The authors divide the observed spectrum and the template spectrum respectively into several parts, and measure the correlations of the corresponding parts of them, which is chosen as the similarity evidences in the proposed similarity measure. The principle of setting the weights is that the higher the correlation, the higher the corresponding weight. The proposed approach is compared with the method based on spectral line matching and the traditional cross correlation technique by experiments, the results show that the proposed method has a higher correct rate.

[A method for redshift determination of quasars based on cross correlation].

This paper presents a novel method for redshift determination of quasars. Firstly, a group of redshifts were determined using the emission line info extracted from the observed spectrum; Secondly, the template was redshifted according to the candidates, and the correlation between the observed spectrum and the redshifted template was measured. Finally, the redshift candidate corresponding to the highest correlation was chosen as the redshift. Compared with the existing methods based on spectral line matching, the proposed method has a lower dependence on the quality of spectral line extraction. Experiments show that this method is robust and superior to the methods based on spectral linematching.

[Mean shift based auto-extraction of spectral lines for non-emission-line objects].

The mean shift algorithm is used. At first, the property that mean shift vectors always point toward local maxima of the density is used to get the pseudo continuum; secondly, mean shift filtering is a goodedge preserving smoothing, which canadaptively reduce the amount of smoothing near feature spectral lines, so the authors use mean shift filtering in noise reduction after the noramalization of continuum spectra; finally, the authors extract feature spectral lines by setting local thresholds. The experiments on both stars and normal galaxies show that our method can extract spectral lines accurately, which is helpful to the parameter measure and the automatic classification of spectra based on spectral lines.

[Density estimation based model matching method for redshift determination].

The present paper proposes a model matching method based on density estimation for redshift determination, in whichthe problem of redshift determination is translated into the problem of searching for the point of maximum density within a data set. At first, the mean shift-based method for auto-extraction of spectral lines is used to get feature spectrallines. Secondly, according tothe redshift formula, the authors use the feature wavelength array and the spectral template to get a data set. Finally, the authors findthe point of maximum density within the data set, then the average of the data in epsilon-neighbor of the point is regarded as the redshift estimation. The information of feature wavelength and spectral line type is used in this method so that it can deal with every kind of spectra. Experiments show that our method is stable and the correct identification rate is high.

Automatic blastomere recognition from a single embryo image.

The number of blastomeres of human day 3 embryos is one of the most important criteria for evaluating embryo viability. However, due to the transparency and overlap of blastomeres, it is a challenge to recognize blastomeres automatically using a single embryo image. This study proposes an approach based on least square curve fitting (LSCF) for automatic blastomere recognition from a single image. First, combining edge detection, deletion of multiple connected points, and dilation and erosion, an effective preprocessing method was designed to obtain part of blastomere edges that were singly connected. Next, an automatic recognition method for blastomeres was proposed using least square circle fitting. This algorithm was tested on 381 embryo microscopic images obtained from the eight-cell period, and the results were compared with those provided by experts. Embryos were recognized with a 0 error rate occupancy of 21.59%, and the ratio of embryos in which the false recognition number was less than or equal to 2 was 83.16%. This experiment demonstrated that our method could efficiently and rapidly recognize the number of blastomeres from a single embryo image without the need to reconstruct the three-dimensional model of the blastomeres first; this method is simple and efficient.