Extension of higher order local autocorrelation features

doi:10.1016/j.patcog.2006.10.006

Pattern Recognition

Volume 40, Issue 5, May 2007, Pages 1466-1473

https://doi.org/10.1016/j.patcog.2006.10.006 Get rights and content

Abstract

This study investigates effective image features that are widely applicable in image analysis. We specifically address higher order local autocorrelation (HLAC) features, which are used in various applications. The original HLAC features are restricted up to the second order and are represented by 25 mask patterns. We increase their orders up to eight and extract the extended HLAC features using 223 mask patterns. Furthermore, we create large mask patterns and construct multi-resolution features to support large displacement regions. In texture classification and face recognition, the proposed method outperformed Gaussian Markov random fields, Gabor features, and local binary pattern operator.

Introduction

Computer image analysis has been investigated actively for many years. To date, it has produced widespread practical applications. Recent great advances in computer technology are further expanding its fields of application. That trend increases the fundamental need for basic image features that are applicable to various tasks. Many basic features such as gray-level histograms, Fourier features and Gabor features have been proposed. Among them, we specifically address higher order local autocorrelation (HLAC) features [1] because they offer many advantages.

One advantage of HLAC features is their wide availability for image analysis. They have demonstrated good performance in face recognition [2], [3], natural object recognition [4], gesture recognition [5], etc. Another advantage that they offer is feature representation using mask patterns. Various mask patterns are easily producible without limitations imposed in terms of specific directions or symmetric distributions. Using them, we can characterize complex distributions of image intensities. The use of mask patterns for feature calculation is suitable for digital image processing (no need of additional calculation such as interpolation). Their use allows simple and fast calculation of HLAC features. Furthermore, they are suitable for hardware implementation. A vision chip that calculates HLAC features has been built. It allows even faster processing [6]. These properties are all desirable for practical applications.

The original HLAC features are restricted up to the second order and are represented by 25 mask patterns. We increase their orders up to eight and represent the extended HLAC features using 223 mask patterns. Each feature value corresponds to the power spectrum of the mask pattern. Therefore, the use of many mask patterns allows detailed characterization of an image. We also use large mask patterns to enlarge the support regions. Using different mask sizes together, we can construct multi-resolution HLAC features.

We evaluate the performance of the proposed method through several texture classifications using images of different sizes, scaled images, and rotated images. Results obtained in the area of texture classification will suggest their availability in other applications such as medical image analysis, remote sensing, and segmentation of scene images. To examine their applicability to an object with shape, we also conduct face recognition. The performance of the proposed method is compared with that obtained using other popular methods: Gaussian Markov random fields (GMRF) [7], Gabor features [8], and local binary pattern (LBP) operator [9].

Section snippets

Conventional HLAC features

The Nth-order autocorrelation functions, extensions of autocorrelation functions, are defined as $x (a_{1}, a_{2}, \dots, a_{N}) = \int f (r) f (r + a_{1}) \dots f (r + a_{N}) d r,$ where $f (r)$ denotes the intensity at the observing pixel r, and $a_{1}, a_{2}, \dots, a_{N}$ are N displacements. HLAC features are primitive image features based on Eq. (1) [1]. Their orders and displacements are arbitrary. However, high-order features with a large displacement region become extremely numerous. Hence, the original HLAC features are restricted up to the second order

Texture classification

We conducted five tests of gray-scale texture classification to evaluate the performance of the proposed method. The “Outex database” [11] was used. It is an empirical evaluation framework for texture analysis that contains widely various texture images and precisely specified classification problems. Table 1 presents a description of the experimental data. In each Test, the performance was evaluated by classifying the test samples that have no overlap with the training samples.

We compared our

Face recognition: AT&T database

In the experiments described above, we applied the proposed method to texture classifications in which the images were characterized based on homogeneous properties. Subsequently, as described in this section, we applied it to face recognition to show that the proposed method is also effective for objects with shape. The conventional second-order HLAC features have demonstrated good performance in face recognition [2], [3]. For that reason, the extended method is also expected to perform well.

Conclusions

We proposed an extension of the HLAC features, which are widely available in image analysis. We increased their orders up to eight and represented the extended HLAC features with 223 mask patterns. By this extension, the image is characterized more closely than conventional second-order features. We also constructed multi-resolution HLAC features that contain both high-frequency and low-frequency information. In texture classification and face recognition, the proposed method demonstrated its

References (21)

M. Kreutz et al.
Scale-invariant image recognition based on higher order autocorrelation features
Pattern Recognition
(1996)
N. Otsu, T. Kurita, A new scheme for practical flexible and intelligent vision systems, in: Proceedings of the IAPR...
T. Kurita, N. Otsu, T. Sato, A face recognition method using higher order local autocorrelation and multivariate...
F. Goudail et al.
Face recognition system using local autocorrelations and multiscale integration
IEEE Trans. Pattern Anal. Mach. Intell.
(1996)
T. Kurita, S. Hayamizu, Gesture recognition using HLAC features of PARCOR images and HMM based recognizer, in:...
K. Yamamoto, I. Ishii, A design of higher order auto-correlation vision chip, IEICE Trans. Inf. Sys. (D-II) J86-D-II(8)...
R. Chellappa et al.
Classification of textures using Gaussian Markov random fields
IEEE Trans. Acoust. Speech Signal Process.
(1985)
B.S. Manjunath et al.
Texture features for browsing and retrieval of image data
IEEE Trans. Pattern Anal. Mach. Intell.
(1996)
T. Ojala et al.
Multiresolution gray-scale and rotation-invariant texture classification with local binary patterns
IEEE Trans. Pattern Anal. Mach. Intell.
(2002)
V. Popovici, J.P. Thiran, Higher order autocorrelations for pattern classification, in: Proceedings of the...

There are more references available in the full text version of this article.

Cited by (14)

Image representation for generic object recognition using higher-order local autocorrelation features on posterior probability images
2012, Pattern Recognition
This paper presents a novel image representation method for generic object recognition by using higher-order local autocorrelations on posterior probability images. The proposed method is an extension of the bag-of-features approach to posterior probability images. The standard bag-of-features approach is approximately thought of as a method that classifies an image to a category whose sum of posterior probabilities on a posterior probability image is maximum. However, by using local autocorrelations of posterior probability images, the proposed method extracts richer information than the standard bag-of-features. Experimental results reveal that the proposed method exhibits higher classification performances than the standard bag-of-features method.
Interest filter vs. interest operator: Face recognition using Fisher linear discriminant based on interest filter representation
2008, Pattern Recognition Letters
This paper introduces a novel Fisher discriminant classifier based on the interest filter representation for face recognition. Our interest Fisher classifier (IFC), which is robust to illumination and facial expression variability, applies the Fisher linear discriminant (FLD) to an augmented interest feature vector derived from interest filter representation of face images. The novelty of this paper comes from our proposed interest filter: the interest operator can reveal the local activity of the images but suffer from some drawbacks and we improve the capability of the interest operator and propose a multi-orientation and multi-scale interest filter. In particular, we carry out comparative studies of different similarity measures applied to various classifiers. We also perform comparative experimental studies of various face recognition schemes, including our novel IFC method, the Eigenfaces and the Fisherfaces methods, the combination of interest operator and the Eigenfaces method, the combination of interest operator and the Fisherfaces method, the Eigenfaces on the augmented interest feature vectors and other popular subspace methods. The feasibility of the new IFC method has been successfully tested on two data sets from the FERET and AR databases. The novel IFC method achieves the highest accuracy on face recognition on both two datasets.
Content-based image retrieval of cultural heritage symbols by interaction of visual perspectives
2011, International Journal of Pattern Recognition and Artificial Intelligence
Study of statistical methods for texture analysis and their modern evolutions
2020, Engineering Reports
Patch Autocorrelation Features: a translation and rotation invariant approach for image classification
2018, Artificial Intelligence Review
Laughing voice recognition using periodic waveforms and voice-likeness features
2016, 2016 IEEE International Conference on Robotics and Biomimetics, ROBIO 2016

View all citing articles on Scopus

View full text

Extension of higher order local autocorrelation features

Abstract

Introduction

Section snippets

Conventional HLAC features

Texture classification

Face recognition: AT&T database

Conclusions

Pattern Recognition

Face recognition system using local autocorrelations and multiscale integration

IEEE Trans. Pattern Anal. Mach. Intell.

Classification of textures using Gaussian Markov random fields

IEEE Trans. Acoust. Speech Signal Process.

Texture features for browsing and retrieval of image data

IEEE Trans. Pattern Anal. Mach. Intell.

Multiresolution gray-scale and rotation-invariant texture classification with local binary patterns

IEEE Trans. Pattern Anal. Mach. Intell.