Academia.eduAcademia.edu

Outline

Title
Abstract
Figures
Introduction
Related Work
Conclusions
References
All Topics
Computer Science
Computer Vision

Image Classification Using Humps of Histogram

Profile image of Adarsh K KhareAdarsh K Khare
https://doi.org/10.1109/CICT.2016.37
visibility

…

description

9 pages

link

1 file

Abstract

Classification techniques classify the remotely sensed image by using reflectance properties of pixels. This paper presents a new approach to classify multispectral remotely sensed image. This approach classifies the multispectral image using frequencies of spectral bands' grey level values (DN values) in Histogram. It draws histogram for different spectral bands of the image. Then, it finds and separates the humps in histograms. This approach yields more meaningful classification for multi-modal or bi-modal histograms. It creates 3 potential centroids in each hump for each spectral band. More the number of humps, more would be potential centroids for classification. Different spectral bands have different peaks in their humps of histograms. It reads all the pixels of one peak of one band and draw the local histogram of other bands' grey level values using pixels read. This way, peak of one hump of one band can find corresponding peaks in local histogram and these peaks make a pixel that can be a potential centroid and some of these peak frequencies is the actual frequency of that centroid. Now, I choose extreme left and extreme right grey level values whose frequency is greater than or equal to the average frequency of that hump. As each hump of each spectral band has three grey level values, I can find three centroids for each hump of each spectral band. Duplicate centroids are eliminated from the list of centroids. The rest of the centroids are recursively iterated and centroids with lesser frequencies than the nearby centroids are eliminated. Later, algorithm uses gravitational force to find out two nearby centroids.

Figures (25)
Fig. 2. Process flow diagram for classifying the image using humps of histogram. Fig. 1. Sample four single band images of different spectral bands. This sample image is used for classification in this paper.
Fig. 2. Process flow diagram for classifying the image using humps of histogram. Fig. 1. Sample four single band images of different spectral bands. This sample image is used for classification in this paper.
diagram of my approach to classify the remotely sensec image.
diagram of my approach to classify the remotely sensec image.
Two Humps: PEAK FREQUENCIES FoR BOTH THE HUMPS
Two Humps: PEAK FREQUENCIES FoR BOTH THE HUMPS
In this stage, the starting line of histogram is raised to find out the disjoint humps programmatically. Here, I use threshold frequency of 3000 to make humps disjoint.
In this stage, the starting line of histogram is raised to find out the disjoint humps programmatically. Here, I use threshold frequency of 3000 to make humps disjoint.
BAND 1: PEAK FREQUENCIES FoR HuMP-1 AND Hump-2
BAND 1: PEAK FREQUENCIES FoR HuMP-1 AND Hump-2
Here, in hump1, band 2 has peak frequency of 5587 for grey level value 57 and peak frequency of 13764 for grey level value 145. Now, while reading all the pixels of multispectral
Here, in hump1, band 2 has peak frequency of 5587 for grey level value 57 and peak frequency of 13764 for grey level value 145. Now, while reading all the pixels of multispectral
Fig. 5. Classified image using humps’ peaks’ grey level values as centroids (2 humps as 2 clusters)
Fig. 5. Classified image using humps’ peaks’ grey level values as centroids (2 humps as 2 clusters)
Fig. 6. Histogram for peaks of band-1 in different humps Now, I can draw another histogram for band-1 in different humps where band-1 grey level values are 87 (for hump-1) and 140 (for hump-2).
Fig. 6. Histogram for peaks of band-1 in different humps Now, I can draw another histogram for band-1 in different humps where band-1 grey level values are 87 (for hump-1) and 140 (for hump-2).
TABLE V 4) Band-4 Humps Here, in hump1, band 3 has peak frequency of 4896 for grey level value 48 and peak frequency of 8000 for grey level value 187. Now, while reading all the pixels of multispectral image, I can see the values of other spectral bands on the same co- ordinate where spectral band 3 has grey level value of 48. Here band-3 frequency of 4896 is global frequency (found after reading whole image) while corresponding band-1, band-
TABLE V 4) Band-4 Humps Here, in hump1, band 3 has peak frequency of 4896 for grey level value 48 and peak frequency of 8000 for grey level value 187. Now, while reading all the pixels of multispectral image, I can see the values of other spectral bands on the same co- ordinate where spectral band 3 has grey level value of 48. Here band-3 frequency of 4896 is global frequency (found after reading whole image) while corresponding band-1, band-
BAND 4: PEAK FREQUENCIES FoR HuMP-1 AND Hump-2 Here, in hump1, band 4 has peak frequency of 6922 for grey level value 29 and peak frequency of 9673 for grey level value 144. Now, while reading all the pixels of multispectral image, I can see the values of other spectral bands on the same co-
BAND 4: PEAK FREQUENCIES FoR HuMP-1 AND Hump-2 Here, in hump1, band 4 has peak frequency of 6922 for grey level value 29 and peak frequency of 9673 for grey level value 144. Now, while reading all the pixels of multispectral image, I can see the values of other spectral bands on the same co-
Fig. 7. Histogram for peaks of band-2 in different humps Now, I can draw another histogram for band-2 in different humps where band 2 grey level values are 57 (for hump-1) and 145 (for hump-2).
Fig. 7. Histogram for peaks of band-2 in different humps Now, I can draw another histogram for band-2 in different humps where band 2 grey level values are 57 (for hump-1) and 145 (for hump-2).
Fig. 8. Histogram for peaks of band-3 in different humps
Fig. 8. Histogram for peaks of band-3 in different humps
INCLUSION OF EXTREME LEFT AND EXTREME RIGHT
INCLUSION OF EXTREME LEFT AND EXTREME RIGHT
D. Finding All the Centroids and Classifying
D. Finding All the Centroids and Classifying
Fig. 10. Classified image using 8 centroids of humps’ peaks. If classify the image using all these 8 centroids, the classified image look like Fig. 10.
Fig. 10. Classified image using 8 centroids of humps’ peaks. If classify the image using all these 8 centroids, the classified image look like Fig. 10.
Two Humps: PEAK FREQUENCIES FOR BOTH HUMPS Fig. 12 shows the peak, extreme left and extreme right frequencies with black line in histogram. Table IX shows the list of extreme left and extreme right centroids of all the humps.
Two Humps: PEAK FREQUENCIES FOR BOTH HUMPS Fig. 12 shows the peak, extreme left and extreme right frequencies with black line in histogram. Table IX shows the list of extreme left and extreme right centroids of all the humps.
In Table X Hump-1, Centroid {144, 150, 208, and 162} is duplicated. Similarly in Table XI Hump-2, Centroid {90, 60, 50, and 30} is triplicated. All the duplicated and triplicated centroids are eliminated and only one copy with highest sum of frequency is retained.
In Table X Hump-1, Centroid {144, 150, 208, and 162} is duplicated. Similarly in Table XI Hump-2, Centroid {90, 60, 50, and 30} is triplicated. All the duplicated and triplicated centroids are eliminated and only one copy with highest sum of frequency is retained.
Similarly average frequencies for all the bands in minor (1*) hump are 4614, 4468, 3910, and 5172 respectively, and average frequencies for all the bands in major (2"’) hump are 11664, 9473, 6397, and 7408 respectively. The average minor hump frequency is 4541 and average major hump frequency is 8735. Fig. 12. Histogram showing extreme left, peak, and extreme right band values’ frequencies in black line.
Similarly average frequencies for all the bands in minor (1*) hump are 4614, 4468, 3910, and 5172 respectively, and average frequencies for all the bands in major (2"’) hump are 11664, 9473, 6397, and 7408 respectively. The average minor hump frequency is 4541 and average major hump frequency is 8735. Fig. 12. Histogram showing extreme left, peak, and extreme right band values’ frequencies in black line.
Here width of minor humps are 7, 6, 6, and 5 respectively, while width of major humps are 28, 37, 59, and 51 respectively. Average width of minor (1") hump is 6, and average width of major (2"') hump is 44.
Here width of minor humps are 7, 6, 6, and 5 respectively, while width of major humps are 28, 37, 59, and 51 respectively. Average width of minor (1") hump is 6, and average width of major (2"') hump is 44.

Related papers

Classification of Multispectral Images using Symbolic Gravitational Approach
D.S. Vinod

National Conference in Recent Trends in Information Technology, 2002

Multispectral image data are well adopted for monitoring and analyzing the behaviour of different covers and also the temporal changes occurring on the target. Periodically collecting the data and then comparing it with the previous data can observe this. Using various sensors, data is collected and further analyzed to obtain information about objects and areas under investigation. The number of features in each sample of the multispectral data depends upon the number of channels from which the information is collected. Data from different channels are combined to obtain the detailed information about the body or area under observation. Here the multispectral data is used to classify, Satellite and Biomedical images. We propose an improvised method of classification of such data set using gravitational symbolic clustering. Symbolic data is a special case of conventional data, which is closer to real life interpretation and analysis. The multispectral image is a quantitative interval type of symbolic data. It requires lot of time and memory space to analyze or classify the data. In order to overcome such limitations a data reduction technique is employed. The data reduction technique uses bin arrays to store useful information of the reduced image. The major idea is to present a clustering algorithm based on gravitational approach. The procedure is based on the physical phenomenon in which a system of particles in space converge to the centroid of the system due to the gravitational attraction between the particles. The concept of mutual pairs is used to merge the samples. The process of merging reduces the number of samples each time they are available for consideration. The process terminates at some time when no more mutual pairs are available for merging. A detailed study is carried on different sets of multispectral images and results are presented. (D.S. Vinod and T.N. Nagabhushana, Classification of Multispectral Images using Symbolic Gravitational Approach, National Conference in Recent Trends in Information Technology, page 29 to 34, held at Karpagam Arts and Science College, Coimbatore, India on 22nd to 24th of August 2002. Organized and published by Department of Computer Science, Karpagam Arts and Science College, Coimbatore, India.)

View PDFchevron_right
Distance histogram to centroid as a unique feature to recognize objects
RAFAEL MOLINA CARMONA

The shape of objects plays an essential role among the different aspects of visual information. A 2D silhouette often conveys enough information to allow the correct recognition of the original 3D object. Distance Histogram to Centroid will be used as the unique feature to totally describe an object and to distinguish it from all the other objects in the scene. The proposed system has been proved to be robust to discriminate between classes in a given set of objects The main advantages are the elimination of the feature selection process and avoiding the problem of dimensionality.

View PDFchevron_right
Homogeneity Distance Classification Algorithm (HDCA): A Novel Algorithm for Satellite Image Classification
Mohammad Karimi Firozjaei

Remote Sensing, 2019

Image classification is one of the most common methods of information extraction from satellite images. In this paper, a novel algorithm for image classification based on gravity theory was developed, which was called “homogeneity distance classification algorithm (HDCA)”. The proposed HDCA used texture and spectral information for classifying images in two iterative supplementary computing stages: (1) merging, (2) traveling and escaping operators. The HDCA was equipped by a new concept of distance, the weighted Manhattan distance (WMD). Moreover, an improved gravitational search algorithm (IGSA) was applied for selecting features and determining optimal feature space scale in HDCA. In the case of multispectral satellite image classification, the proposed method was compared with two well-known classification methods, Maximum Likelihood classifier (MLC) and Support Vector Machine (SVM). The results of the comparison indicated that overall accuracy values for HDCA, MLC, and SVM are 9...

View PDFchevron_right
Optimization strategies for rapid centroid estimation
Steven Su

Abstract²Particle swarm algorithm has been extensively utilized as a tool to solve optimization problems. Recently proposed particle swarm±based clustering algorithm called the Rapid Centroid Estimation (RCE) is a lightweight alteration to Particle Swarm Clustering (PSC). The RCE in its standard form is shown to be superior to conventional PSC algorithm. We have observed some limitations in RCE including the possibility to stagnate at a local minimum combination and the restriction in swarm size. We propose strategies to optimize RCE further by introducing RCE+ and swarm RCE+. Five benchmark datasets from UCI machine learning database are used to test the performance of these new strategies. In Glass dataset swarm RCE+ is able to achieve highest purity centroid combinations with less iteration (90.3%±1.1% in 9±5 iterations) followed by RCE+ (89%±3.5% in 65±62 iterations) and RCE (87%±5.9% in 54±44). Similar quality is also reflected in other benchmark datasets including Iris, Wine, Breast Cancer, and Diabetes.

View PDFchevron_right
Hyperspectral Image Classification Using a Self-Organizing Map
Pablo Martínez

2000

View PDFchevron_right
Gravitational clustering: a new approach based on the spatial distribution of the points
Sukhamay Kundu

Pattern Recognition, 1999

View PDFchevron_right
Identification of Geospatial Objects Using Spectral Pattern
Subhabrata Barman

Remote Sensing Techniques and GIS Applications in Earth and Environmental Studies, 2017

Solar radiation on hitting a target surface may be transmitted, absorbed or reflected. Different materials reflect and absorb differently at different wavelengths. The reflectance spectrum of a material is a plot of the fraction of radiation reflected as a function of the incident wavelength and serves as a unique signature for the material. In principle, a material can be identified from its spectral reflectance signature if the sensing system has sufficient spectral resolution to distinguish its spectrum from those of other materials. This premise provides the basis for multispectral remote sensing. Nguyen Dinh Duong (1997) proposed a method for decomposition of multispectral image into several sub-images based on modulation (spectral pattern) of the spectral reflectance curve. The hypothesis roots from the fact that different ground objects have different spectral reflectance and absorption characteristics which are stable for a given sensor. This spectral pattern can be considered as invariant and be used as one of classification rules.

View PDFchevron_right
Processing Multispectral Scanner Data Using Correlation Clustering and Nonparametric Classification Techniques
Richard E Haskell

1976

View PDFchevron_right
Pattern recognition with generalized centroids and subcentroids
Shoude Chang

Applied Optics, 2005

We propose a class of generalized moment functions (GMFs) that can be used to determine a set of geometric points, namely, generalized centroids (G centroids), within an object. Based on a linear GMF, a mass centroid and its subcentroids can be defined and extracted, which provide information on the location and orientation of an object. Similar to traditional moment functions, GMFs can also be used to describe the global shape of an object, including symmetry and fullness. However, GMFs, along with G centroids and subcentroids, can further serve to construct a feature vector of an object, which is critical for image registration and invariant pattern recognition. One can extract more distinguishing features from the same object by changing the combination of GMFs. We present results that show simulations of pattern recognition from uniform backgrounds.

View PDFchevron_right
Image Segmentation Based on a Two-Dimensional Histogram
Mohamed El Ansari

Image Segmentation, 2011

View PDFchevron_right

References (21)

  1. Mengqiu Tian, Qiao Yang, Andreas Maier, Ingo Schasiepen, Nicole Maass1, Matthias Elter, "An automatic histogram-based initializing algorithm for K-means clustering in CT", Proceedings des Workshops Bildverarbeitung für die Medizin 2013 (Bildverarbeitung für die Medizin 2013), Heidelberg, Germany, 03.03.2013, pp. 277-282, 2013
  2. J. Puzicha, T. Hofmann, and J. Buhmann, "Histogram Clustering for Un- supervised Segmentation and image retrival", Pattern Recognition Letters, 20(9):899-909, 1999
  3. "Hierarchical cluster algorithm for remote sensing data of earth", V. S. Sidorova,, Representation, Processing, Analysis, And Understanding Of Images, Pattern Recognition and Image Analysis, June 2012, Volume 22, Issue 2, pp 373-379
  4. Shu-Ling Shieh, Tsu-Chun Lin, and Yu-Chin Szu, "An Efficient Clustering Algorithm Based on Histogram Threshold", Intelligent Information and Database Systems Volume 7197 of the series Lecture Notes in Computer Science pp 32-39
  5. Alireza Khotanzad and Abdelmajid Bouarfa, "Image Segmentation By A Parallel, Non-Parametric Histogram Based Clustering Algorithm", Pattern Recogmition, Vol. 23, No. 9, pp. 961-973
  6. V. S. Sidorova, "Automatic hierarchical clustering algorithm for remote sensing data", Representation, Processing, Analysis And Understanding Of Images Pattern Recognition and Image Analysis 21, Issue 2, pp 328-331
  7. Patel, R., Shrawankar, U.N., Raghuwanshi, M.M with Histogram Construction Technique. In: Proc Second International Conference on Emerging Tren Technology, pp. 615-618. IEEE Computer Society
  8. Xiuping Jia, Member, IEEE, and John A. Rich Representation for Hyperspectral Data Clas Transactions on GeoScience and Remote Sensing, V 2002
  9. Biplab Banerjee, B. Krishna Mohan, "A Novel Clustering Technique for Uunspervised Classi Sensing Images", ISPRS Technical Commission V 12 December 2014, Hyderabad, India.
  10. Maas, S. J. and Rajan, N. (2010). Normalizing and data using scatter plot matching. Remote Sensing, 2
  11. S.M. Ali, "New Fully Automatic Multispectral based on Scatterplot Method", ISSN 2250-24
  12. Certified Journal, Volume 3, Issue 10, October 201
  13. Ms. Chandrakala.M , Mrs. R.Amsaveni, "Class Sensing Image Areas Using Surf Features a Allocation", ijarcsse Volume 3, Issue 9, September
  14. Teodor Costachioiu, Rodica Constantinescu, Bash Datcu, "Semantic Analysis Of Satellite Image European Signal Processing Conference (EUSIPC Romania, August 27 -31, 2012.
  15. D. Landgrebe, "Hyperspectral Image Data A Dimensional Signal Processing Problem" IEEE Magazine, Vol 19, No. 1 pp. 17-28, January 2002.
  16. F. Hammadi-Mesmoudi and J. Korczak, "An network classifier and its application in remote sen International Conference on Image Processing, (Ed
  17. A. Ketterlin, D. Blamont, and J. Korczak, "Unsu spatial regularities," in Proc. of the European Sym Remote Sensing, SPIE Proc. Series 2579, 1995.
  18. N. R. Pal and S. Pal, "A review on image segm Pattern Recognition 26, pp. 1277-1294, 1993.
  19. T. Kohonen, "The self-organizing map," in Proc 1464-1480, 1990.
  20. R. M. Haralick and L. Shapiro, "Image segme Computer Vision, Graphics, and Image Processin 1985.
  21. T. Kohonen, Self-Organization and Associative Me in Information Sciences 8, Heidelberg, 1984.

Related papers

NUnsupervised Classification Using Gravity Centers from Scatter Plot
Adarsh K Khare

Classification techniques classify the remotely sensed image by using reflectance properties of pixels. This paper presents a new approach to classify multispectral remotely sensed image. This approach classifies the multispectral image using frequencies of spectral bands' grey level values (DN values) in Histogram. It draws histogram for different spectral bands of the image. Then, it finds and separates the humps in histograms. This approach yields more meaningful classification for multi-modal or bi-modal histograms. It creates 3 potential centroids in each hump for each spectral band. More the number of humps, more would be potential centroids for classification. Different spectral bands have different peaks in their humps of histograms. It reads all the pixels of one peak of one band and draw the local histogram of other bands' grey level values using pixels read. This way, peak of one hump of one band can find corresponding peaks in local histogram and these peaks make a pixel that can be a potential centroid and some of these peak frequencies is the actual frequency of that centroid. Now, I choose extreme left and extreme right grey level values whose frequency is greater than or equal to the average frequency of that hump. As each hump of each spectral band has three grey level values, I can find three centroids for each hump of each spectral band. Duplicate centroids are eliminated from the list of centroids. The rest of the centroids are recursively iterated and centroids with lesser frequencies than the nearby centroids are eliminated. Later, algorithm uses gravitational force to find out two nearby centroids.

View PDFchevron_right
Unsupervised Classification Using Gravity Centers from Scatter Plot
Adarsh K Khare

— Unsupervised classification creates clusters by grouping pixels based on the reflectance properties of pixels. This paper presents a new approach to classify multi-spectral remotely sensed image using pixels' density in N-dimensional scatterplot. It first finds the densely populated clusters in N-dimensional scatter plot and then finds gravity centers of these densely populated clusters. Later, multispectral image is classified using minimum distance to gravity classifier. At the beginning of classification, this approach neither makes extreme assumption of considering each pixel as a different cluster nor goes to the other extreme by considering all the pixels in a single cluster. It follows the middle path by making assumption of some pixels as gravity centers of different clusters before classifying the image. It creates clusters of equal size in N-dimensional scatter plot and picks up the densely populated clusters. All these clusters are recursively iterated for self-adjustment of gravity centers using mathematical algorithm. The approach uses gravitational force for merging two nearby clusters. Here, gravity center of a cluster is calculated by summing up the spectral bands' values and dividing it by number of pixels within that cluster. When two nearby clusters are merged, their gravity centers are also adjusted accordingly. This provides the most densely populated clusters and their gravity centers. Now, these gravity centers can be used to classify the remotely sensed image using minimum distance to gravity center classifier.

View PDFchevron_right
Classification of Remotely Sensed data using Gravitational Symbolic Clustering
D.S. Vinod

Remotely sensed data is obtained through artificial satellites from various geographic and astronomical sources. Remotely sensed image data are well adopted for monitoring and analyzing the behavior of different covers and also the temporal changes occurring on the target. Observations are made by periodically collecting the data and then comparing previous data. Using various sensors, data is collected and further analyzed to obtain information about objects and areas under investigation. The number of features in each sample of the 'Remotely sensed data' or 'multispectral image' depends upon the number of channels from which the information is collected. Data from different channels are combined to obtain the detailed information about the body or area under observation. We propose an improvised method of classification of such data using gravitational symbolic clustering. Symbolic data is a special case of conventional data, which is closer to real life interpretation and analysis. In conventional data, the objects are individualized, whereas in symbolic data sets they are more unified by means of relationships. The multispectral image is a quantitative interval type of symbolic data. It requires lot of time and memory space to analyze or classify the data. In order to overcome such limitations a data reduction technique is proposed. The data reduction technique uses storage bin arrays to store useful information of the reduced image data. The major idea is to present a clustering algorithm based on gravitational symbolic approach. The procedure is based on the physical phenomenon in which a system of particles in space converges to the centroid of the system due to the gravitational attraction between the particles. Both Agglomerative and Disaggregative Gravitational Symbolic approaches are considered. The concept of mutual pairs is used to merge the samples. The process of merging reduces the number of samples each time they are available for consideration. The process terminates at a stage when no more mutual pairs are available for merging. A detailed study has been carried out on different sets of multispectral images. We have also compared our results using cluster indices.

View PDFchevron_right
HINDU: Histogram Inspired Neighborhood Discerning Unsupervised Pattern Recognition System
Belur Dasarathy

1976

View PDFchevron_right
Classification of Satellite Images using New Fuzzy Cluster Centroid for Unsupervised Classification Algorithm
Heltin Genitha C, K. Vani

Vagueness in the boundaries of land cover classes is one of the important problems in the image classification. Fuzzy c means (FCM) is a traditional clustering algorithm that has been widely used in the satellite image classification. However, this algorithm has the drawback of falling into a local minimum and it needs much time to accomplish the classification for a large data set. In order to overcome these drawbacks, a New Fuzzy Cluster Centroid (NFCC) for unsupervised classification algorithm is proposed to improve the traditional FCM and fuzzy weighted c means (FWCM) algorithm. In this work a, new objective function is formulated by adding the new term along with the distance between the pixels and cluster centers in the spectral domain. This new term is formulated by multiplying the Lagrange’s multiplier with the membership values of the pixel for a particular class is subtracted with one. It gives weightage to the instance of a particular pixel. The inclusion of the fuzzy centroid for each cluster increases the stability of the algorithm and the inclusion of the new term reduces the number of iterations for image classification. The technique was applied to both IKONOS and QuickBird images. Overall accuracy statistics indicates that NFCC for unsupervised classification algorithm increases the accuracy of the satellite image classification at the pixel level. Keywords: Fuzzy c-means, IKONOS images, QuickBird images, classification, local minimum.

View PDFchevron_right

Related topics

  • Image Processing
  • Graph based semi-supervised lear...
  • Supervised Visitation
    • 580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved