Multicolor fluorescence in situ hybridization (M-FISH) techniques provide color karyotyping that allows simultaneous analysis of numerical and structural abnormalities of whole human chromosomes. Chromosomes are stained combinatorially in M-FISH. By analyzing the intensity combinations of each pixel, all chromosome pixels in an image are classified. Due to the overlap of excitation and emission spectra and the broad sensitivity of image sensors, the obtained images contain crosstalk between the color channels. The crosstalk complicates both visual and automatic image analysis and may eventually affect the classification accuracy in M-FISH. The removal of crosstalk is possible by finding the color compensation matrix, which quantifies the color spillover between channels. However, there exists no simple method of finding the color compensation matrix from multichannel fluorescence images whose specimens are combinatorially hybridized. In this paper, we present a method of calculating the color compensation matrix for multichannel fluorescence images whose specimens are combinatorially stained.

Multicolor fluorescence in situ hybridization (M-FISH) techniques provide color karyotyping that allows simultaneous analysis of numerical and structural abnormalities of whole human chromosomes. Chromosomes are stained combinatorially in M-FISH. By analyzing the intensity combinations of each pixel, all chromosome pixels in an image are classified. Often, the intensity distributions between different images are found to be considerably different and the difference becomes the source of misclassifications of the pixels. Improved pixel classification accuracy is the most important task to ensure the success of the M-FISH technique. In this paper, we introduce a new feature normalization method for M-FISH images that reduces the difference in the feature distributions among different images using the expectation maximization (EM) algorithm. We also introduce a new unsupervised, nonparametric classification method for M-FISH images. The performance of the classifier is as accurate as the maximum-likelihood classifier, whose accuracy also significantly improved after the EM normalization. We would expect that any classifier will likely produce an improved classification accuracy following the EM normalization. Since the developed classification method does not require training data, it is highly convenient when ground truth does not exist. A significant improvement was achieved on the pixel classification accuracy after the new feature normalization. Indeed, the overall pixel classification accuracy improved by 20% after EM normalization.

In this paper, a novel scheme is introduced for human facial feature extraction. Unlike previous methods that fit a 3D morphable model to 2D intensity images, our scheme utilizes 3D range images to extract features without requiring manually-defined initial landmark points. A linear transformation is used to achieve the mapping between the 3D model and a 3D range image, which makes the computation simple and fast. Moreover, our scheme is robust to the illumination and pose variations. In addition to features from range images, extra features can be obtained by examining optional 2D texture images. Using our scheme, we can also perform automatic eye/mouth corner localization. Experimental results show the high accuracy and robustness of our scheme.

Interest in face recognition systems has increased significantly due to the emergence of significant commercial opportunities in surveillance and security applications. In this paper we propose a novel technique to extract features from 3D face representations. In this technique, first the nose tip is automatically located on the range image, then the range data from a hexagonal region of interest around this landmark is decomposed using Barycentric wavelet kernels. The dimensionality of the extracted coefficients at each resolution level is reduced using principal component analysis (PCA). These new features are tested on 206 range images, and a high classification accuracy is achieved using a small number of features. The obtained accuracy is competitive to that of other techniques in literature.

In this paper, we describe a novel 3-D face recognition scheme for 3-D face recognition that can automatically identify faces from range images, and is insensitive to holes, facial expression, and hair. In our scheme, a number of carefully selected range images constitute a set of example faces, and another range image is chosen as a ldquogeneric face.rdquo The generic face is then warped to match each of the example faces in the least mean square sense. Each such warp is specified by a vector of displacement values. In feature extraction operation, when a target face image comes in, the generic face is warped to match it. The geometric transformation used in the warping is a linear combination of the example face warping vectors. The coefficients in the linear combination are adjusted to minimize the root mean square error. After the matching process is complete, the coefficients of the composite warp are used as features and passed to a Mahalanobis-distance-based classifier for face recognition. Our technique is tested on a data set containing more than 600 range images. Experimental results in the access-control scenario show the effectiveness of the extracted features.

Biomedical specimens thicker than the microscope depth-of-field (DOF) produces images of limited quality where a part of the specimen is in focus and others out of focus. Even the best in-focus image always contains objects that are blurred. These blurred contents make the quantitative and accurate analysis of the specimen difficult. The problem can be remedied by extending the DOF of the imaging system. One way to achieve this is to apply digital image fusion, a software technique that combines multiple images into a carefully synthesized image. Another way, albeit more expensive, is to incorporate modification in the optical system. In this paper we describe an extended DOF method that is based on multiscale pointwise product (MPP) image fusion, and compare it with the state-of-the-art wavefront coding technique which involves optical modification. We show competitive performance of our method and contrast its pros and cons with those of the wavefront coding.

Multicolor fluorescence in-situ hybridization (M-FISH) technique provides color karyotyping that allows simultaneous analysis of numerical and structural abnormalities of whole human chromosomes. Currently available M-FISH systems exhibit misclassifications of multiple pixel regions that are often larger than the actual chromosomal rearrangement. This paper presents a novel unsupervised classification method based on fuzzy logic classification and a prior adjusted reclassification method. Utilizing the chromosome boundaries, the initial classification results improved significantly after the prior adjusted reclassification while keeping the translocations intact. This paper also presents a new segmentation method that combines both spectral and edge information. Ten M-FISH images from a publicly available database were used to test our methods. The segmentation accuracy was more than 98% on average.

Two novel non-ubiquitous digital watermarking schemes are proposed and evaluated in parallel (one in the spatial domain and one in the wavelet domain) for the purpose of record indexing and integrity protection of medical images. A fingerprint of the regions of interest (ROI) in a medical image is generated using the MD5 message-digest algorithm [R. Rivest, April 1992], and it is embedded along with the medical record information into the non-ROI parts of the image. Our experiments show that with both schemes, a payload of 420 bits of information can be achieved watermarking clinical cytogenetic images, without affecting image content in the ROI nor incurring perceptible differences between original and watermarked images. Furthermore, the resulting watermarked images can survive common image processing operations such as cropping, JPEG compression, sharpening, and their combinations.

Chromosomes are essential genomic information carriers. Chromosome classification constitutes an important part of routine clinical and cancer cytogenetics analysis. Cytogeneticists perform visual interpretation of banded chromosome images according to the diagrammatic models of various chromosome types known as the ideograms, which mimic artists' depiction of the chromosomes. In this paper, we present a subspace-based approach for automated prototyping and classification of chromosome images. We show that 1) prototype chromosome images can be quantitatively synthesized from a subspace to objectively represent the chromosome images of a given type or population, and 2) the transformation coefficients (or projected coordinate values of sample chromosomes) in the subspace can be utilized as the extracted feature measurements for classification purposes. We examine in particular the formation of three well-known subspaces, namely the ones derived from principal component analysis (PCA), Fisher's linear discriminant analysis, and the discrete cosine transform (DCT). These subspaces are implemented and evaluated for prototyping two-dimensional (2-D) images and for classification of both 2-D images and one-dimensional profiles of chromosomes. Experimental results show that previously unseen prototype chromosome images of high visual quality can be synthesized using the proposed subspace-based method, and that PCA and the DCT significantly outperform the well-known benchmark technique of weighted density distribution functions in classifying 2-D chromosome images.

This paper presents Microarray BASICA: an integrated image processing tool for background adjustment, segmentation, image compression and analysis of microarray images. BASICA uses the fast Mann-Whitney test-based algorithm introduced in J. Hua et al., (2002) to segment microarray images, and post-processing to eliminate the segmentation irregularities. The segmentation results, along with the foreground and background intensities obtained with background adjustment, are then used for the independent compression of foreground and background. We introduce a new distortion measure for microarray image compression and devise a coding scheme by modifying the object-based embedded block coding with optimized truncation (object-based EBCOT) algorithm J. Hua et al., (2002) to achieve optimal rate-distortion performance in lossy coding while still maintaining outstanding lossless compression performance. Experimental results show that BASICA can extract sufficiently accurate genetic information at bitrates as low as 43bpp.

Chromosome banding patterns are very important features for karyotyping, based on which cytogenetic diagnosis procedures are conducted. Due to cell culture, staining, and imaging conditions, image enhancement is a desirable preprocessing step before performing chromosome classification. In this paper, we apply a family of differential wavelet transforms (Wang and Lee, 1998), (Wang, 1999) for this purpose. The proposed differential filters facilitate the extraction of multiscale geometric features of chromosome images. Moreover, desirable fast computation can be realized. We study the behavior of both banding edge pattern and noise in the wavelet transform domain. Based on the fact that image geometrical features like edges are correlated across different scales in the wavelet representation, a multiscale point-wise product (MPP) is used to characterize the correlation of the image features in the scale-space. A novel algorithm is proposed for the enhancement of banding patterns in a chromosome image. In order to compare objectively the performance of the proposed algorithm against several existing image-enhancement techniques, a quantitative criteria, the contrast improvement ratio (CIR), has been adopted to evaluate the enhancement results. The experimental results indicate that the proposed method consistently outperforms existing techniques in terms of the CIR measure, as well as in visual effect. The effect of enhancement on cytogenetic diagnosis is further investigated by classification tests conducted prior to and following the chromosome image enhancement. In comparison with conventional techniques, the proposed method leads to better classification results, thereby benefiting the subsequent cytogenetic diagnosis.

Image interpolation has been extensively studied. In this paper, we present a fast filter bank algorithm for interpolating images at any scales. The proposed algorithm is a generalization of [1] and [2], which handles only either integer or dyadic scale magnification of images. The details of the algorithm and implementation are described in the paper. In particular, we show that several image interpolation algorithms are the special cases of our proposed algorithm. The comparison with other interpolating bases are also discussed. The proposed algorithm is of practical use for enlargement in display and printing applications.

On the subject of optimal subspaces for appearance-based object recognition, it is generally believed that algorithms based on LDA (linear discriminant analysis) are superior to those based on PCA (principal components analysis), provided that relatively large training data sets are available. In this paper, we show that while this is generally true for classification with the nearest-neighbor classifier, it is not always the case with a maximum-likelihood classifier. We support our claim by presenting both intuitively plausible arguments and actual results on a large data set of human chromosomes. Our conjecture is that perhaps only when the underlying object classes are linearly separable would LDA be truly superior to other known subspaces of equal dimensionality.

Differential operators have been widely used for multiscale geometric descriptions of images. Efficient computation of these differential operators can be obtained by taking advantage of the spline techniques. We make use of a special class of these operators for image enhancement, with a particular application to chromosome image enhancement. These operators constitute a translation invariant wavelet transform well suited for the structural description of chromosome geometry. Based on the fact that the geometrical features like edges are correlated between different scales in the representation, a novel algorithm is designed to enhance the salient features of the image. Comparisons of this algorithm with other approaches are presented.

We describe a novel method of enhancing image geometric features using the oriented wavelets introduced in our earlier work (see Yu-Ping Wang, IEEE Trans. Image Proc., vol.8, no.12, p.1757-71, 1999). The poor directional selectivity of the conventional 2D wavelet transform has been circumvented by using this class of oriented wavelets. By taking advantage of directional wavelet decomposition, both the directional and scale correlation information are utilized for enhancing salient structures in images. To remove phase dependence, a pair of quadrature filters are used. The proposed algorithm has been applied to chromosome image enhancement. Comparisons of this algorithm with other approaches are presented.

For over forty years, two seemingly unrelated fields have formed an unlikely partnership due to a subtle symbiosis. These two fields are cytogenetics and digital imaging. Cytogenetics is the biological study of the genetic makeup of cells. It is used extensively in genetic diagnosis and cancer research. Digital imaging is the computer engineering technique of processing pictures in a computer. It is widely used in medicine, military, industrial, entertainment, and aerospace applications. Digital imaging has contributed to cytogenetics instrumentation that reduces the workload in clinical labs and produces quantitative data for both research and diagnosis. At the same time, cytogenetics has furnished the problems that have prompted the development of a broad range of techniques that prove useful in fields far beyond genetics. This talk traces the parallel and interacting histories of these two fields. Breakthroughs in one have led to advances in the other throughout their development. This illustrates how necessity in one field is the mother of invention in another, and how both have benefited from their interaction.

Automated chromosome classification has been an important pattern recognition problem for decades. Numerous attempts were made in the past to characterize chromosome band patterns as part of the feature description vector. In this paper we describe a recent study to employ wavelet packets as basis function sets to compute chromosome band pattern features. A total of 28 wavelet packet basis function sets were evaluated in this study. The experimental results are presented and compared with those currently best-performing method on two benchmark chromosome datasets.