A bivariate autoregressive technique for analysis and classification of planar shapes

Das, M.; Paulik, Mark; Loh, N.K.

doi:10.1109/34.41389

Cited by 52 publications

(10 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Examples include Das et al, 20 Dubois and Glanz, 21 Eom, 22 He and Kundu, 23 and Kashyap and Chellappa. 24 There are other shape-analysis techniques (e.g., Gunsel and Telkap 25 and Neil and Curtis 26 ) that fall out side of these classes, and some that combine features of two or more classes (e.g., Cohen et al and Wang et al 28 ).…”

Section: A Taxonomy Of Techniquesmentioning

confidence: 97%

Comparing Photometric Distributions

Ashdown

2000

Journal of the Illuminating Engineering Society

View full text Add to dashboard Cite

Section: A Taxonomy Of Techniquesmentioning

confidence: 97%

Comparing Photometric Distributions

Ashdown

2000

Journal of the Illuminating Engineering Society

View full text Add to dashboard Cite

“…100% correct classification results are reported in [3] and [4]. The major disadvantage is that the schemes are very sensitive to shape occlusion, even slight occlusion.…”

Section: Introductionmentioning

confidence: 99%

Shape classification using hidden Markov model

Kundu

1991

[Proceedings] ICASSP 91: 1991 International Conference on Acoustics, Speech, and Signal Processing

View full text Add to dashboard Cite

In this paper, we present a planar shape recognition approach based on hidden Markov model and autoregressive parameters. This approach segments closed shapes into segments and explores the characteristic relations between consecutive segments to make classification at finer level. The algorithm can tolerate a lot of shape contour perturbation and moderate amount of occlusion. Also, the overall classification scheme is independent of shape orientation. Excellent recognition result has been reported. A distinct advantage of the approach is that the classifier does not have to be trained all over again when a new class of shapes is added. I. I N T R O D U C T I O NTwo dimensional shape classification is an important problem in computer vision and image processing. There currently exist many shape classification techniques. Some of the earlier techniques are summarized in 111. They include Fourier descriptors, the class of space domain or curve fitting methods featuring the chain code approach, the polygonal approximation scheme, the category of decomposition technique including the medial axis transform (MAT), decomposition at concave vertices, decomposition by clustering, and the scale domain representation. All of these methods have their advantages and limitations.Another very effective technique is the Autoregressive (AR) model approach [2-41. 100% correct classification results are reported in [3] and [4]. The major disadvantage is that the schemes are very sensitive to shape occlusion, even slight occlusion. In this case, only 35% to 70% recognition rates were reported. The schemes are also sensitive to shape contour distortion or perturbation.The reason for this is that these schemes model the whole shape with only one set of predictive parameters. If the shape contains a large number of sample points, and the contour varies radically, the shape may seem rather unpredictable. In this case, it is reasonable to think that an AR This work has been supported in part by National Science Foundation Grant MIP-8908082 model with finite number of parameters is not adequate for the whole shape. This problem inspires the consideration of another important technique in pattern recognitionthe hidden Markov model.Hidden Markov Model (HMM) explores the relationship between consecutive segments of a pattern to be classified. Each segment is relatively smaller, and therefore is easier to be characterized. HMM has been used successfully in many applications. For example, Rabiner e l a1 have applied HMM to speech recognition 15). Kundu e t a1 have developed a handwritten word recognition scheme using HMM [SI. In these applications, better results are achieved compared to other methods that treat the patterns as a single set of features.In this paper, we will present a technique combining HMM and AR model to recognize closed 2-D shapes. Our approach is to segment the 1-D representation sequence of a closed shape into several pieces, characterize each piece with AR parameters and get a vector sequence for each shape, and then ap...

show abstract

“…Contour-based descriptors [9], [16], [20], [21], [22] are useful in both pattern recognition and data compression. Kauppinen et al [9] and Sekita et al [16] compared the performance of several such descriptors for recognizing 2D patterns.…”

Section: Contour-based Fourier Descriptormentioning

confidence: 99%

Hidden Markov models with spectral features for 2D shape recognition

Cai

Liu

2001

IEEE Trans. Pattern Anal. Machine Intell.

View full text Add to dashboard Cite

AbstractÐIn this paper, we present a technique using Markov models with spectral features for recognizing 2D shapes. We will analyze the properties of Fourier spectral features derived from closed contours of 2D shapes and use these features for 2D pattern recognition. We develop algorithms for reestimating parameters of hidden Markov models. To demonstrate the effectiveness of our models, we have tested our methods on two image databases: hand-tools and unconstrained handwritten numerals. We are able to achieve high recognition rates of 99.4 percent and 96.7 percent without rejection on these two sets of image data, respectively.

show abstract

A bivariate autoregressive technique for analysis and classification of planar shapes

Cited by 52 publications

References 22 publications

Comparing Photometric Distributions

Comparing Photometric Distributions

Shape classification using hidden Markov model

Hidden Markov models with spectral features for 2D shape recognition

Contact Info

Product

Resources

About