Pengwei Hao

Matrix factorizations for reversible integer mapping

Reversible integer mapping is essential for lossless source coding by transformation. A general matrix factorization theory for reversible integer mapping of invertible linear transforms is developed. Concepts of the integer factor and the elementary reversible matrix (ERM) for integer mapping are introduced, and two forms of ERM-triangular ERM (TERM) and single-row ERM (SERM)-are studied. We prove that there exist some approaches to factorize a matrix into TERMs or SERMs if the transform is invertible and in a finite-dimensional space. The advantages of the integer implementations of an invertible linear transform are (i) mapping integers to integers, (ii) perfect reconstruction, and (iii) in-place calculation. We find that besides a possible permutation matrix, the TERM factorization of an N-by-N nonsingular matrix has at most three TERMs, and its SERM factorization has at most N+1 SERMs. The elementary structure of ERM transforms is the ladder structure. An executable factorization algorithm is also presented. Then, the computational complexity is compared, and some optimization approaches are proposed. The error bounds of the integer implementations are estimated as well. Finally, three ERM factorization examples of DFT, DCT, and DWT are given.

Compound image compression for real-time computer screen image transmission

We present a compound image compression algorithm for real-time applications of computer screen image transmission. It is called shape primitive extraction and coding (SPEC). Real-time image transmission requires that the compression algorithm should not only achieve high compression ratio, but also have low complexity and provide excellent visual quality. SPEC first segments a compound image into text/graphics pixels and pictorial pixels, and then compresses the text/graphics pixels with a new lossless coding algorithm and the pictorial pixels with the standard lossy JPEG, respectively. The segmentation first classifies image blocks into picture and text/graphics blocks by thresholding the number of colors of each block, then extracts shape primitives of text/graphics from picture blocks. Dynamic color palette that tracks recent text/graphics colors is used to separate small shape primitives of text/graphics from pictorial pixels. Shape primitives are also extracted from text/graphics blocks. All shape primitives from both block types are losslessly compressed by using a combined shape-based and palette-based coding algorithm. Then, the losslessly coded bitstream is fed into a LZW coder. Experimental results show that the SPEC has very low complexity and provides visually lossless quality while keeping competitive compression ratios.

Fingerprint indexing based on composite set of reduced SIFT features

Most of current fingerprint indexing schemes utilize features based on global textures and minutiae structures. To extend the existing technology of feature extraction, this paper proposes a new fingerprint indexing and retrieval scheme using scale invariant feature transformation (SIFT), which has been widely used in generic image retrieval. With slight loss in effectiveness, we reduce the number of features generated from one fingerprint for efficiency. To cope with the uncertainty of acquisition (e.g. partialness, distortion), we use a composite set of features to form multiple impressions for the fingerprint representation. In the index construction phase, the use of locality-sensitive hashing (LSH) allows us to perform similarity queries by only examining a small fraction of the database. Experiments on database FVC2000 and FVC2002 show the effectiveness of our proposed scheme.

Reversible integer KLT for progressive-to-lossless compression of multiple component images

In this paper, we presented a method for integer reversible implementation of KLT for multiple component image compression. The progressive-to-lossless compression algorithm employed the JPEG-2000 transform coding strategy using the multiple component transform (MCT) across the components, followed by a 2-dimensional wavelet transform on individual eigen images. The linear MCTs we tested and compared are KLT, discrete wavelet transform (DWT), and a tasselled cap transform (TCT) for TM satellite images only. The computational complexity of the reversible integer implementation is no more than that of naive transformation, and the overhead data is very small. Its effectiveness was evaluated using two 6-band landsat TM satellite images and an 80-component hyper-spectral remotely-sensed image. Experiments with KLT and wavelet based JPEG-2000 show that reversible KLT (RKLT) outperforms other approaches for all of the test images in the case of both lossy and lossless compression.

Comparative study of color transforms for image coding and derivation of integer reversible color transform

A color transform is necessary for better color image coding. In this paper, we concentrate on a comparative study of color transforms for color image coding in order to find the best one among 11 published color transforms: YCrCb, NTSC, PAL, HDTV, WW, XYZ, DCT, DHT, two approximate K-L transforms (K1K2K3 and KLT) and the original reversible color transform (ORCT) adopted in JPEG-2000. Experiment results with JPEG-2000 verification model (VM5.1) are sorted by color transforms and counted up for all the test images and diverse bit rates. The sorting scoring table shows that the discrete cosine transform performs best among 11 tested color transform at 6 lossy bit rates for all the 2 JPEG-2000 color test images and other 23 color images we used. Then we derive an integer reversible transform of DCT and an approximate implementation using additions and shifts only for both lossless and lossy color image coding. Experiments with the integer reversible color transforms show that the proposed transform scheme is better than ORCT for lossy image coding.

Fingerprint indexing based on singular point correlation

Fingerprint indexing is an efficient technique that greatly improves the performance of automated fingerprint identification systems. We propose a continuous fingerprint indexing method based on location, direction estimation and correlation of fingerprint singular points. Location and direction estimation are achieved simultaneously by applying a T-shape model to directional field of fingerprint images. The T-shape model analyzes homocentric sectors around the candidate singular points to find lateral-axes and further main-axes. Then a distortion-tolerant filter of minimum average correlation energy is utilized to obtain a correlation-based similarity measure which gives the evidence of searching priority. The experiment is performed by 400-fingerprint retrieval from 10,000 templates and the mean search space is only 3.46% of the whole dataset.

Integer reversible transformation to make JPEG lossless

JPEG, as an international image coding standard based on DCT and Huffman entropy coder, is still popular in image compression applications although it is lossy. JPEG-LS, standardized for lossless image compression, however, employs an encoding technique different from JPEG. This paper presents an integer reversible implementation to make JPEG lossless. It uses the framework of JPEG, and just converts DCT and color transform to be integer reversible. Integer DCT is implemented by factoring the float DCT matrix into a series of elementary reversible matrices and each of them is directly integer reversible. Our integer DCT integrates lossy and lossless schemes nicely, and it supports both lossy and lossless compression by the same method. Our JPEG can be used as a replacement for the standard JPEG in either encoding or decoding or both. Experiments show that the performance of JPEG with our integer reversible DCT is very close to that of the original standard JPEG for lossy image coding, and more importantly, with our transform, it can compress images losslessly.

A Geometric Method for Optimal Design of Color Filter Arrays

A color filter array (CFA) used in a digital camera is a mosaic of spectrally selective filters, which allows only one color component to be sensed at each pixel. The missing two components of each pixel have to be estimated by methods known as demosaicking. The demosaicking algorithm and the CFA design are crucial for the quality of the output images. In this paper, we present a CFA design methodology in the frequency domain. The frequency structure, which is shown to be just the symbolic DFT of the CFA pattern (one period of the CFA), is introduced to represent images sampled with any rectangular CFAs in the frequency domain. Based on the frequency structure, the CFA design involves the solution of a constrained optimization problem that aims at minimizing the demosaicking error. To decrease the number of parameters and speed up the parameter searching, the optimization problem is reformulated as the selection of geometric points on the boundary of a convex polygon or the surface of a convex polyhedron. Using our methodology, several new CFA patterns are found, which outperform the currently commercialized and published ones. Experiments demonstrate the effectiveness of our CFA design methodology and the superiority of our new CFA patterns.

Hallucinating faces from thermal infrared images

This paper addresses the face hallucination problem of converting thermal infrared face images into photo-realistic ones. It is a challenging task because the two modalities are of dramatical difference, which makes many developed linear models inapplicable. We propose a learning-based framework synthesizing the normal face from the infrared input. Compared to the previous work, we further exploit the local linearity in not only the image spatial domain but also the image manifolds. We have also developed a measurement of the variance between an input and its prediction, thus we can apply the Markov random field model to the predicted normal face to improve the hallucination result. Experimental results show the advantage of our algorithm over the existing methods. Our algorithm can be readily generalized to solve other multi-modal image conversion problems as well.

New color filter arrays of high light sensitivity and high demosaicking performance

For high light sensitivity, new CFA designs use panchromatic pixels, aka white pixels, that no visible spectrum energy is filtered. Kodaks CFA2.0 has 50% white pixels, but the demosaicking performance is not good. We present in this work a set of new color filter arrays (CFA) of high light sensitivity and high demosaicking performance which were obtained by using a CFA design methodology in the frequency domain. The new patterns are of size 5×5 and come from the same frequency structure, which has one luma in the base band at (0, 0) and four chromas (two conjugate pairs) placed at (4π/5, 2π/5), (— 4π/5, — 2π/5), (2π/5, — 4π/5) and (— 2π/5, 4π/5), respectively. The new patterns are optimized to have only white (panchromatic) and three primary color pixels and the pixels are found to be 40% white, 20% red, 20% green and 20% blue by pixel color constrained optimization. Our demosaicking experiments show that our new CFA patterns outperform Kodak CFA2.0 in both objective and subjective quality.