Rotation Invariant Pattern Recognition Research Articles

Image analysis by log-polar Exponent-Fourier moments

Moments, as a popular class of the global invariant image descriptors, have been widely used in image analysis, pattern recognition and computer vision applications. Exponent-Fourier moments (EFMs) are a new set of orthogonal moments based on exponential functions, which are suitable for image analysis and rotation invariant pattern recognition. However, EFMs lack natively the scaling-invariant property. In addition, they always suffer from high time complexity, numerical instability, and reconstruction error, especially for higher order of moments. In this paper, we introduce a class of scaling and rotation-invariant orthogonal moments, named Log-Polar Exponent-Fourier moments (LPEFMs), by extending the classical EFMs to the log-polar coordinates. Firstly, we redefined the EFMs’ basis functions in log-polar domain instead of Cartesian/polar coordinate domain in order to obtain the scaling-invariant property. Then, we develop a new framework for computing the LPEFMs by using pseudo-polar Fourier transform and frequency domain interpolation, which result in better image representation capability, numerical stability, and computational speed. Compared with the classical EFMs, the proposed LPEFMs have four advantages, scaling invariance, speed, accuracy and stability. Theoretical analysis and simulation results are provided to validate the proposed image moment and to compare its performance with previous works.

The modified generic polar harmonic transforms for image representation

This paper introduces four classes of orthogonal transforms by modifying the generic polar harmonic transforms. Then, the rotation invariant feature of the proposed transforms is investigated. Compared with the traditional generic polar harmonic transforms, the proposed transforms have the ability to describe the central region of the image with a parameter controlling the area of the region. Experimental results verified the image representation capability of the proposed transforms and showed better performance of the proposed transform in terms of rotation invariant pattern recognition.

Phase-Only Rotation Invariant Correlation Using Synthesized Phase Objects.

We present a development of the method of synthesized phase objects (SPO-method) (P. V. Yezhov, et al. Opt. Exp. 20, 29854 (2012)) for the phase-only rotation invariant pattern recognition. It has been performed a comparison of correlation signals for a set of amplitude objects under their rotation by using the standard and SPO methods, by applying the Fourier-Mellin transformation. The results of both calculation and optical experiments carried out using an optical-digital correlator with SLM in the Fourier plane have been presented.

Position and rotation-invariant pattern recognition system by binary rings masks

In this paper, algorithms invariant to position, rotation, noise and non-homogeneous illumination are presented. Here, several manners are studied to generate binary rings mask filters and the corresponding signatures associated to each image. Also, in this work it is shown that digital systems, which are based on the -law non-linear correlation, are -invariant for . The methodologies are tested using greyscale fossil diatoms digital images (real images), and considering the great similarity between those images the results obtained are excellent. The box plot statistical analysis and the computational cost times yield that the Bessel rings masks are the best option when the images contain a homogeneous illumination and the Fourier masks digital system is the right selection when the non-homogeneous illumination and noise is presented in the images.

Hough-transform based algorithm for the automatic invariant recognition of rectangular chocolates. Detection of defective pieces.

La Transformada de Hough (TH) ha sido usada para crear un algoritmo de reconocimiento invariante (rotación, traslación y cambio de escala) de objetos poligonales. En esta investigación, el sistema desarrollado ha sido exitosamente aplicado al reconocimiento automático invariante de chocolates rectangulares, incluyendo la identificación de piezas falladas. Como la ejecución total del sistema resultante es totalmente autónoma, los resultados pueden ser aplicados al control automático de calidad en aplicaciones industriales.

Orthogonal moments based on exponent functions: Exponent-Fourier moments

In this paper, we propose a new set of orthogonal moments based on Exponent functions, named Exponent-Fourier moments (EFMs), which are suitable for image analysis and rotation invariant pattern recognition. Compared with Zernike polynomials of the same degree, the new radial functions have more zeros, and these zeros are evenly distributed, this property make EFMs have strong ability in describing image. Unlike Zernike moments, the kernel of computation of EFMs is extremely simple. Theoretical and experimental results show that Exponent-Fourier moments perform very well in terms of image reconstruction capability and invariant recognition accuracy in noise-free, noisy and smooth distortion conditions. The Exponent-Fourier moments can be thought of as generalized orthogonal complex moments.

Rotation-Invariant Neural Pattern Recognition System Using Extracted Descriptive Symmetrical Patterns

In this paper a novel rotation-invariant neural-based pattern recognition system is proposed. The system incorporates a new image preprocessing technique to extract rotation-invariant descriptive patterns from the shapes. The proposed system applies a three phase algorithm on the shape image to extract the rotation-invariant pattern. First, the orientation angle of the shape is calculated using a newly developed shape orientation technique. The technique is effective, computationally inexpensive and can be applied to shapes with several non-equally separated axes of symmetry. A simple method to calculate the average angle of the shape’s axes of symmetry is defined. In this technique, only the first moment of inertia is considered to reduce the computational cost. In the second phase, the image is rotated using a simple rotation technique to adapt its orientation angle to any specific reference angle. Finally in the third phase, the image preprocessor creates a symmetrical pattern about the axis with the calculated orientation angle and the perpendicular axis on it. Performing this operation in both the neural network training and application phases, ensures that the test rotated patterns will enter the network in the same position as in the training. Three different approaches were used to create the symmetrical patterns from the shapes. Experimental results indicate that the proposed approach is very effective and provide a recognition rate up to 99.5%.

Robust Partial Shape Recognition Using Curvature and an Iterative Closest Point Algorithm

This article presents a robust translation, rotation, and scaling invariant pattern recognition algorithm using partially distorted two-dimensional (2D) contours. Both global and local matching accuracies are guaranteed: the former is achieved by using global control point (GCP) and curvature matching and the latter by using the iterative closest point (ICP) algorithm. In the proposed algorithm, an input contour is considered two possible types: first the contour contains GCP and the second contains no GCP. For the contours with no detected GCPs, the algorithm considers the contour as the second type of contour and the curvature matching algorithm is directly applied. For the contour with GCPs, a global alignment is performed using the GCPs and then, the curvature matching algorithm is applied. The validity of the algorithm is illustrated by the presentation of experimental results. The experiments show that the algorithm works well for partial object recognition with and without GCPs.

Image analysis by Bessel–Fourier moments

In this paper, we proposed a new set of moments based on the Bessel function of the first kind, named Bessel–Fourier moments (BFMs), which are more suitable than orthogonal Fourier–Mellin and Zernike moments for image analysis and rotation invariant pattern recognition. Compared with orthogonal Fourier–Mellin and Zernike polynomials of the same degree, the new orthogonal radial polynomials have more zeros, and these zeros are more evenly distributed. The Bessel–Fourier moments can be thought of as generalized orthogonalized complex moments. Theoretical and experimental results show that the Bessel–Fourier moments perform better than the orthogonal Fourier–Mellin and Zernike moments (OFMMs and ZMs) in terms of image reconstruction capability and invariant recognition accuracy in noise-free, noisy and smooth distortion conditions.

Content-addressable holographic data storage system for invariant pattern recognition of gray-scale images

Conventionally a holographic data storage system uses binary digital data as the input pages. We propose and demonstrate the use of a holographic data storage system for the purpose of invariant pattern recognition of gray-scale images. To improve the correlation accuracy for gray-scale images, we present a coding technique, phase Fourier transform (phase-FT) coding, to code a gray-scale image into a random and balanced digital binary image. In addition to the fact that a digital data page is obtained for incorporation into a holographic data storage system, this phase-FT coded image produces dc-free homogenized Fourier spectrum. This coded image can also be treated as an image for further processing, such as synthesis of distortion-invariant filters for invariant pattern recognition. A space-domain synthetic discriminant function (SDF) filter has been synthesized using these phase-FT coded images for rotation-invariant pattern recognition. Both simulation and experimental results are presented. The results show good correlation accuracy in comparison to correlation results obtained for SDF filter synthesized using the original gray-scale images themselves.

Rotation-invariant pattern recognition using morphological fringe-adjusted joint transform correlation

We propose a rotation-invariant nonlinear correlator based on circular harmonic filter (CHF) and morphological Fringe-adjusted Joint Transform correlation (MFJTC). We refer to this correlator as a rotation-invariant MFJTC (RIMFJTC). Through computer simulation, we compare the output results of RIMFJTC with those of rotation-invariant MC (RIMC) and CHF when input scene is corrupted by salt-and-pepper noise, white additive Gaussian noise and cluttered background. Our results show that RIMFJTC yields higher discriminability, sharper and higher correlation peaks, and displays better stability against the above three kinds of noise than do the RIMC and common CHF.

INVARIANT PATTERN RECOGNITION USING RIDGELET PACKETS AND THE FOURIER TRANSFORM

In this paper, we propose two novel invariant algorithms for pattern recognition by using ridgelet packets and the Fourier transform. Ridgelet packets provide many orthonormal bases that can effectively capture directional features present in pattern images. The Fourier transform is good at eliminating the orientation differences. By combining these two tools, very efficient rotation invariant pattern recognition techniques are created. Experimental results show that the proposed methods achieve very high classification rates and they outperform other state-of-the-art methods for rotation invariant pattern recognition under both noise-free and noisy environments.

Binary phase only filters for rotation and scale invariant pattern recognition with the joint transform correlator

The joint transform correlator (JTC) is one of the two main optical image processing architecture which provides a highly effective way of comparing images in a wide range of applications. Traditionally, an optical correlator is used to compare an unknown input scene with a pre-captured reference image library, to detect if the reference occurs within the input. Strength of the correlation signal decreases rapidly as the input object rotates or varies in scale relative to the reference object. The aim of this paper is to overcome the intolerance of the JTC to rotation and scale changes in the target image. Many JTC systems are constructed with the use of ferroelectric liquid crystal (FLC) spatial light modulators (SLMs) as they provide fast two-dimensional binary modulation of coherent light. Due to the binary nature of the FLC SLMs used in the JTC systems, any image addressed to the device need to have some form of thresholding. Carefully thresholding the grey scale input plane and the joint power spectrum (JPS) has significant effect on the quality of correlation peaks and zero order (DC) noise. A new thresholding technique to binarise the JPS has been developed and implemented optically. This algorithm selectively enhances the desirable fringes in the JPS which provide correlation peaks of higher intensity. Zero order noise is further reduced when compared to existing thresholding techniques. Keeping in mind the architecture of the JTC and limitations of FLC SLMs, a new technique to design rotation and scale invariant binary phase only filters for the JTC architecture is presented. Filers design with this technique have limited dynamic range, higher discriminability among target and non-target objects, and convenience for implementation on FLC SLMs. Simulation and experiments shows excellent results of various rotation and scale invariant filters designed with this technique. A rotation invariant filter is needed for various machine vision applications of the JTC. By fixing the distance between camera and input object, the scale sensitivity of the correlator can be avoided. In contrast to the industrial machine vision applications, scale factor is very important factor for the applications of the JTC systems in defence and security. A security system using a scale invariant JTC will be able to detect a target object well in advance and will provide more time to take a decision.

Rotation invariant pattern recognition using ridgelets, wavelet cycle-spinning and Fourier features

In this paper, we propose a rotation-invariant descriptor for pattern recognition by using ridgelets, wavelet cycle-spinning, and the Fourier transform. Ridgelets have been developed recently and have many advantages over wavelets in applications to image processing. However, the current implementation of ridgelets cannot be applied to pattern recognition directly. In order to overcome this problem, we have successfully extracted ridgelet features within the circle surrounding the pattern we are trying to recognize. Wavelet cycle-spinning and Fourier spectrum magnitudes are used to achieve rotation invariance. The main motivation of using ridgelets is that we have a much better tool for the extraction of features based on line singularities as compared to point singularities as in the case of wavelets. Based on this observation, important features can be extracted. Our experiments show that our proposed descriptor is very robust to Gaussian noise and it achieves high recognition rates.

Rotation invariant color pattern recognition by use of a three-dimensional Fourier transform.

Recently, the use of three-dimensional correlation for multichannel pattern recognition has been introduced. In this work we propose the use of circular harmonic components with this new technique to obtain invariance under target rotations. The differences between this method and the previous use of circular harmonic filters for multichannel images are discussed. Also the problem of determining the proper center is studied and, to our knowledge, a new and more understandable criterion to locate it is introduced. Some simulation results to verify the successful operation of the method are included.

Improved rotation invariant pattern recognition using circular harmonics of binary gray level slices

We introduce a new rotation invariant pattern recognition method based on nonlinear correlation. The images are decomposed into disjoint binary slices and then correlated using the common linear correlation. This operation is very discriminant even when the target is embedded in strong noise. We extend our sliced orthogonal nonlinear generalized correlation method to rotation invariant pattern recognition by combining the information of a circular harmonic (CH) of each binary slice of the reference object with binary slices of the target. In addition to improved discrimination capability, the method avoids the time-consuming process of finding proper centers for the CHs. Results are presented including the study of the stability of the correlation peaks in the presence of background noise and overlapping Gaussian noise.

Phase-shifting technique applied to circular harmonic-based joint transform correlator

The phase-shifting technique is applied to the circular harmonic expansion-based joint transform correlator. Computer simulation has shown that the light efficiency and the discrimination capability are greatly enhanced, and the full rotation invariance is preserved after the phase-shifting technique has been used. A rotation-invariant optical pattern recognition with high discrimination capability and high light efficiency is obtained. The influence of the additive noise on the performance of the correlator is also investigated. However, the anti-noise capability of this kind of correlator still needs improving.

Nonlinear rotation-invariant pattern recognition by use of the optical morphological correlation

We introduce a modification of the nonlinear morphological correlation for optical rotation-invariant pattern recognition. The high selectivity of the morphological correlation is conserved compared with standard linear correlation. The operation performs the common morphological correlation by extraction of the information by means of a circular-harmonic component of a reference. In spite of some loss of information good discrimination is obtained, especially for detecting images with a high degree of resemblance. Computer simulations are presented, as well as optical experiments implemented with a joint transform correlator.

Hypothetical neural mechanism that may play a role in mental rotation: an attractor neural network model

We propose an attractor neural network (ANN) model that performs rotation-invariant pattern recognition in such a way that it can account for a neural mechanism being involved in the image transformation accompanying the experience of mental rotation. We compared the performance of our ANN model with the results of the chronometric psychophysical experiments of Cooper and Shepard (Cooper L A and Shepard R N 1973 Visual Information Processing (New York: Academic) pp 204-7) on discrimination of alphanumeric characters presented in various angular departures from their canonical upright position. Comparing the times required for pattern retrieval in its canonical upright position with the reaction times of human subjects, we found agreement in that (i) retrieval times for clockwise and anticlockwise departures of the same angular magnitude (up to 180°) were not different, (ii) retrieval times increased with departure from upright and (iii) increased more sharply as departure from upright approached 180°. The rotation-invariant retrieval of the activity pattern has been accomplished by means of the modified algorithm of Dotsenko (Dotsenko V S 1988 J. Phys. A: Math. Gen. 21 L783-7) proposed for translation-, rotation- and size-invariant pattern recognition, which uses relaxation of neuronal firing thresholds to guide the evolution of the ANN in state space towards the desired memory attractor. The dynamics of neuronal relaxation has been modified for storage and retrieval of low-activity patterns and the original gradient optimization of threshold dynamics has been replaced with optimization by simulated annealing.

ニューラルネットワークによる図形の回転正規化

This paper discusses a neural network for a rotation invariant pattern recognition system. The proposed network can normalize a rotated random pattern into the unrotated standard pattern. The normalization is achieved by cascading a preceding rotation-angle extracting network and the succeedingnormalizing network. This basic structure is the same as the previously reported shifted-pattern normalizing network. The network is trained to normalize input random pattern direction into its center of gravity downwardsor in some other predetermined direction. The weight distribution viewed from the input layer in each hidden unit reveals a Fourier transform like in the radius and angle directions.