Brendan Babb

The best fingerprint compression standard yet

Modern fingerprint compression and reconstruction standards, such as those used by the US Federal Bureau of Investigation FBI), are based upon the 9/7 discrete wavelet transform. This paper describes how a genetic algorithm was used to evolve wavelet and scaling numbers for each level of a multiresolution analysis (MRA) transform that consistently outperforms the 9/7 wavelet for fingerprint compression and reconstruction tasks. Our evolved transforms also improve upon wavelets optimized by a genetic algorithm via the lifting scheme, and thus establish a new state-of-the-art in this important application area.

Evolving optimized matched forward and inverse transform pairs via genetic algorithms

This research established a methodology for using a genetic algorithm to evolve coefficients for matched forward and inverse transform pairs. Beginning with an initial population of randomly mutated copies of the coefficients representing a standard wavelet, our GA consistently evolved transforms that outperformed wavelets for image compression and reconstruction applications under conditions subject to quantization error. Transforms optimized against a single representative image also outperformed wavelets when subsequently tested against other images from our test set. The new methodology has the potential to revolutionize the signal and image processing fields.

Improved multiresolution analysis transforms for satellite image compression and reconstruction using evolution strategies

In this paper, we describe how an evolution strategy optimizes multiresolution analysis (MRA) transforms that outperform wavelets for satellite image compression and reconstruction under conditions subject to quantization error. At three multiresolution levels and 64:1 quantization, our best evolved transform reduces mean squared error (MSE) in reconstructed images by an average of 11.71% (0.54 dB) in comparison to the 9/7 Cohen-Daubechies-Feauveau (CDF) wavelet, while continuing to match the 9/7s compression capabilities. This result establishes a new state-of-the-art for quantized digital satellite images.

Optimized satellite image compression and reconstruction via evolution strategies

This paper describes the automatic discovery, via an Evolution Strategy with Covariance Matrix Adaptation (CMA-ES), of vectors of real-valued coefficients representing matched forward and inverse transforms that outperform the 9/7 Cohen-Daubechies-Feauveau (CDF) discrete wavelet transform (DWT) for satellite image compression and reconstruction under conditions subject to quantization error. The best transform evolved during this study reduces the mean squared error (MSE) present in reconstructed satellite images by an average of 33.78% (1.79 dB), while maintaining the average information entropy (IE) of compressed images at 99.57% in comparison to the wavelet. In addition, this evolved transform achieves 49.88% (3.00 dB) average MSE reduction when tested on 80 images from the FBI fingerprint test set, and 42.35% (2.39 dB) average MSE reduction when tested on a set of 18 digital photographs, while achieving average IE of 104.36% and 100.08%, respectively. These results indicate that our evolved transform greatly improves the quality of reconstructed images without substantial loss of compression capability over a broad range of image classes.

Improved satellite image compression and reconstruction via genetic algorithms

A wide variety of signal and image processing applications, including the US Federal Bureau of Investigations fingerprint compression standard [3] and the JPEG-2000 image compression standard [26], utilize wavelets. This paper describes new research that demonstrates how a genetic algorithm (GA) may be used to evolve transforms that outperform wavelets for satellite image compression and reconstruction under conditions subject to quantization error. The new approach builds upon prior work by simultaneously evolving real-valued coefficients representing matched forward and inverse transform pairs at each of three levels of a multi-resolution analysis (MRA) transform. The training data for this investigation consists of actual satellite photographs of strategic urban areas. Test results show that a dramatic reduction in the error present in reconstructed satellite images may be achieved without sacrificing the compression capabilities of the forward transform. The transforms evolved during this research outperform previous start-of-the-art solutions, which optimized coefficients for the reconstruction transform only. These transforms also outperform wavelets, reducing error by more than 0.76 dB at a quantization level of 64. In addition, transforms trained using representative satellite images do not perform quite as well when subsequently tested against images from other classes (such as fingerprints or portraits). This result suggests that the GA developed for this research is automatically learning to exploit specific attributes common to the class of images represented in the training population.

A differential evolution algorithm for optimizing signal compression and reconstruction transforms

State-of-the-art image compression and reconstruction techniques utilize wavelets. Beginning in 2004, however, a team of researchers at Wright-Patterson Air Force Base (WPAFB), the University of Alaska Anchorage (UAA), and the Air Force Institute of Technology (AFIT) has demonstrated that a genetic algorithm (GA) is capable of evolving non-wavelet transforms that consistently outperform wavelets when applied to a broad class of images under conditions subject to quantization error. Unfortunately, the computational cost of our GA-based approach has been enormous, necessitating hundreds of hours of CPU time, even on supercomputers provided by the Arctic Region Supercomputer Center (ARSC). The purpose of this investigation was to begin to determine whether an alternative approach based upon differential evolution (DE) [20] could be used to (a) optimize transforms capable of outperforming those evolved by the GA, (b) reduce the amount of computation necessary to evolve such transforms, and/or (c) further reduce the mean squared error (MSE) of transforms previously evolved via our GA.

Evolved transforms surpass the FBI wavelet for improved fingerprint compression and reconstruction

This paper describes the evolution of new wavelet and scaling numbers for optimized transforms that consistently outperform the 9/7 discrete wavelet transform (DWT) for fingerprint compression and reconstruction.

Evolved Multiresolution Analysis Transforms for Improved Image Compression and Reconstiruction under Quantization

The research described in this paper uses a genetic algorithm (GA) to evolve wavelet and scaling coefficients for transforms that outperform discrete wavelet transforms (DWTs) under conditions subject to quantization. Compression and reconstruction transform pairs evolved against a representative training image reduce mean squared error (MSE) by more than 22% (1.126 dB) when subsequently applied to test images at a single level of decomposition, while evolved three-level multiresolution analysis (MRA) transforms average more than 11% (0.50 dB) MSE reduction when applied to test images in comparison to the Daubechies-4 (D4) wavelet, without increasing the size of the compressed file

State-of-the-art lossy compression of Martian images via the CMA-ES evolution strategy

The research described in this paper uses the CMA-ES evolution strategy to optimize matched forward and inverse transform pairs for the compression and reconstruction of images transmitted from Mars rovers under conditions subject to quantization error. Our best transforms outperform the 2/6 wavelet (whose integer variant was used onboard the rovers), substantially reducing error in reconstructed images without allowing increases in compressed file size. This result establishes a new state-of-the-art for the lossy compression of images transmitted over the deep-space channel.

Evolving improved transforms for reconstruction of quantized ultrasound images

State-of-the-art lossy compression schemes for medical imagery utilize the 9/7 wavelet. Recent research has established a methodology for using evolutionary computation (EC) to evolve wavelet and scaling numbers describing novel reconstruction transforms that outperform the 9/7 under lossy conditions. This paper describes an investigation into whether evolved transforms could automatically compensate for the detrimental effects of quantization for ultrasound (US) images. Results for 16:1, 32:1, and 64:1 quantization consistently demonstrate superior performance of evolved transforms in comparison to the 9/7 wavelet; in general, this advantage increases in proportion to the selected quantization level.