Implementation of Weighted Structure Similarity & Texture Parameters for Camouflage Texture

Implementation of Weighted Structure Similarity & Texture Parameters for Camouflage Texture

Rahul Patil Surendra Ramteke Nilima Patil

Abstract

Camouflaging is the art of disguising an object to blend the object in a similar background. In this paper method for camouflage texture analysis is proposed. A camouflage texture analysis method based on Weighted Structure Similarity (WSSIM) and Natural Image Parameters between Camouflage texture and background image which are calculated and compared to help to direct the designing camouflage texture. Primary experimental results and comparisons demonstrate that the implemented method is helpful for designing camouflage texture.

1. Introduction

   The term camouflage comes from the French word camoufler meaning to blind. Camouflage also called protective concealment , means to disguise an object in plain sight , in order to conceal it from other environment.

   The simple straight forward requirement in camouflaging an object is to enclose an object in such texture of matching background that human eyes fail to detect the target. In the extended version the target detection even by powerful telescope or remote sensing camera would also be difficult, so it is very important to evaluate whether camouflage texture spots are consistent with the background in the shape, size, color and spatial distribution, traditionally there is limitation and short comings like physiological factors of the observers and testing environment in image quality subjective evaluation.

   In last two decade many objective evaluation

   methods were proposed to assess camouflage texture.

   Sengottuvelan [2] employed the GLCM and Dendogram for natural image to identify the camouflage object. Thomas [4] built up a software system for the camouflage assessment of object in image sequences. Venkata Rao [10] describes an image quality assessment technique which is based on the properties of the human visual system (HVS). It combines the notions of structural similarity with visual regions of interest. Xiaopend Li [3] used Spectrum features as well as texture features of multi-spectral image between target and background, then calculating mahalanobis distance of spectrum features as well as texture features data vector. Nagappa [6] employed co-occurrence matrix to compute texture features within small region of the image. Then defective portions were detected by cluster analysis and identified through watershed segmentation.

   In this paper we implement a weighted structure similarity and natural image features to evaluate the perceived difference between camouflage image texture and background image.

2. Camouflage Texture Evaluation

   A. Weighted Structure Similarity

   A variety of basic operation is performed to eliminate distortion from background image and camouflage texture. And a low-pass filter simulating the point spread function of the eye optics may be applied. Let X = {xi|I = 1,2,,N} and Y = {yi|I = 1,2,,N} be two discrete non-negative signals that have been aligned with each other (e.g., two image patches extracted from the same spatial location from two

   Fig. 1. SSIM measurement system L: low-pass filtering; 2: downsampling by 2.

   images being compared, respectively), and Let µ, and be the mean of X, the variance of X and the covariance of X and Y respectively. Approximately, µ and can be express in terms of the luminance and contrast of X, and measures the tendency of X and Y to vary together, which indicates the structural similarity. In [6] the luminance, contrast and structure comparison measures were given as follows :

   image and the camouflage image signals as the input, the system iteratively applies a low- pass filter and down samples the filtered image by factor of 2. At the jth scale the contrast comparison (2) and the structure comparison

   1. are calculated and denoted as cj(x, y) and sj (x,y), respectively. The luminance comparison (1) is computed only at scale M and is denoted as lM (x,y). The overall SSIM evaluation is obtained by combining the measurement at different scales.

      2 2

      , = 2 +1

      + +2

      (1)

      B) Nature Image Feature Analysis

      2 2

      , = 2 +2

      + +2

      (2)

      Natural image signals are highly structured [13], their pixels exhibit strong dependence which carries important

      , =

      2 +3

      +3

      (3)

      information about the structure objects in visual scene. Structural features of natural images can be used to evaluate the effects of camouflage texture. We select

      some natural image features for directing the design

      Where C1, C2 and C3 are small constants included to avoid instability.

      At last, the structural similarity (SSIM) index between X and Y can be drawn as

      camouflage texture[9]. The selected features are summarized in the following [14]

      • Average Luminance :

        , = , , (, ) (4)

        The camouflage texture is compared with each block of the background with same size to get the whole evaluation result by weighted structure similarity as follows:

        = 1 1

        =

        Where N is the sum of image Pixels.

      • Standard deviation :

   (6)

   1

   = ( 1

   2)1/2 (7)

   , = =1 , (5)

   1

   =1

   Where wj is the weight of different block of background and M is the sum of the blocks.

   The structural similarity measurement system diagram is illustrated in fig. 1. Taking the background

   • Correlation length, which can be convenient from image Fourier spectrum.

   = 10.0/ (8)

   Where,

   2

   2 = 2 +2 |(,)|2

   |(,)|

   (9)

   where F(U,V) with U and V as a variables is image Fourier spectrum.

   • Texture direction, which is the angle of peak of the power spectrum:

     (, ) = | , | (10)

   • Image Entropy :

     =0

     = 1 2() (11)

     Where N is sum of image pixels, L is sum of classes, and C is the pixels number of every class.

     • Edge Detection: By edge detection the number of closed areas and the length of edge can be got. Background image and camouflage textures are treated with canny edge detector to extract the image edges [15].

   So, all these features can measure the difference between camouflage texture and the background image, so according to the differences (e.g. luminance etc.) designers are able to improve the camouflage textures.

3. Experimental Results and Discussion

   The camouflage texture is evaluated based on WSSIM and nature image feature analysis in fig.2 (a) is the background image and Fig.2 (b) (c) and (d) are the camouflage textures for analysis. In Fig.3, the blue box area is compared with the eight neighbor areas. In fig.3

   (a) is the background image (b), (c) and (d) are the camouflage textures. the table I gives test results.

   From table I it is clear that texture 2 is better than texture 1 and texture 3, which is consistent with the subjective evaluation.

   According to these experimental results, designers know whether to acept the camouflage texture and how to improve the camouflage texture.

   1. Background image (b) Camouflage texture1

      ( c) Camouflage texture 2 (d) Camouflage texture3 Figure 2. Background image and Camouflage textures

      1. (b)

   (c) (d)

   Figure 3. Background image and Camouflage textures.

4. Conclusion

   In this paper we implement a WSSIM approach for camouflage texture evaluation and nature image parameter is incorporating the variations of camouflage image resolution and viewing conditions which have the greatest effect on detection. However, this approach is still rather crude. We are working on developing it into a more systematic approach that can potentially be employed in much broader range of application.

   Table 1. Normalized Features Values

   Parameters	Original image	Texture 1	Texture 2	Texture 3
   Deviation of luminance	1	0.0947	0.0956	0.0978
   Image entropy	7.7130	7.7092	7.7137	7.7167
   Correlation length	0.9797	0.9791	0.9797	0.9787
   No. of closed areas	3844	3855	3890	3817
   Index of similarity	1	0.9890	0.9889	0.9888

5. References

1. J. C. Kilian and L. Hepfinger, Computer-based evaluation of camouflage, in Proc.SPIE, vol. 1687, 1992, pp. 358369.

2. P. Sengottuvelan, Amitabh Wahi and A. Shanmugan Performance of Decamouflaging Through Exploratory Image Analysis in IEEE, DOI 10.1109/ICETET 2008.

3. Xiaopend Li Performance Evaluation of Pattern Painting Camouflage based on the Texture Characteristics of Multi- Spectral Images in MIPPR Proc. Of SPIE Vol. MP102 67871Y-1.

4. Thomas Muller and Markus Muller, Computer-aided camouflage assessment in real-time, in Proc.SPIE vol. 6543, 2007, pp. 701711.

5. Thomas Muller and Markus Muller, CACAMO-Computer Aided Camouflage Assessment of Moving Objects, in Proc.SPIE vol. 6941, 2008, pp. 69410V-169410V-12.

6. Nagappa U. Bhajantri and P. Nagabhushan, Camouflage Defect Identification: A Novel Approach, Proceedings of the 9th International Conference on Information Technology, 2006, pp. 145148.

7. R. M. Haralick, K. Shanmugam, and I.Dinstein, Textural Features for Image Classification, IEEE Trans. on Systems, Man and Cybernetics, Vol.SMC-3, No.6, November 1973, pp.610-621.

8. J. S. Weszka, C. R. Dyer, A Comparative Study of Texture Measures for Terrain Classification, IEEE Trans. Syst. Man Cybern, vol. SMC-6, no. 4, 1976.

9. Song Liming and Geng Weidong A New Camouflage Texture Evaluation Method Based on WSSIM and Nature Image Features. IEEE, 978-1-4244-7874, 2010.

10. D.Venkata Rao and L.Pratap Reddy Weighted Structural similarity Based on Edge Strength for Image Quality Assessment, International Journal of Computer Theory and Engineering, Vol. 1, No. 2, June2009. 1793-8201.

11. H. R. Sheikh, A. C. Bovik, L. Cormack, and Z. Wang, LIVE Image Quality Assessment Database,2004, Available On line at http://live.ece.utexas.edu/research/quality

12. VQEG, Final report from the video quality experts group on the validation of objective models of video quality assessment, http://www.vqeg.org/, Mar. 2004.

13. Zhou Wamg, Image Quality Assessment: From Error Visibility to Structural Similarity, IEEE transactions on image processing, Vol.13, No.4, 2004, pp: 600-612.

14. B.Zheng, A.Lu, A.Hardesty, J.H.Sumkin, C.M.Hakim, M.A.Ganott, D.Gur, A method to improve visual similarity of breast masses for an interactive computer-aided diagnosis environment, Medical Physics, Vol.33, No.1, 2006, pp: 111- 117.

15. Kenji Kagitani, Image Quality Evaluation Using Generalized Natural Image, in Proc.SPIE, Vol.6808, 2008, pp: 6808S-1-6808S 8.