Camouflage Breaking A Review of Contemporary Techniques Amy
Camouflage Breaking A Review of Contemporary Techniques Amy Whicker CSCE 867 – Final Project
What is camouflage? The process of masking the foreground to appear as though it is background. Camouflage related work can be divided into two areas: • Camouflage assessment and design • Camouflage breaking • Little has been researched in this area
Why is camouflage breaking important? • Military tactics • Background subtraction • Helps in the understanding of extraction of noncamouflaged objects • Helps in developing algorithm to locates object in the foreground
Camouflage Breaking Methods • Multiple Camouflage Breaking by Co-occurrence and Canny Method developed by: P. Nagabhushan and Nagappa U. Bhajantri • Convexity-based Camouflage Breaking Method developed by: Ariel Tankus and Yehezkel Yeshurun
Co-occurrence and Canny Method Part 1: Determine if there is a camouflaged object in the image. • Create a gray level co-occurrence probability matrix. • Assess the co-occurrence matrix’s texture parameters. Part 2: Achieve effective visualization of camouflage objects. • Repeatedly apply the Canny edge detection operator
Calculating the co-occurrence matrix Example from P. Nagabhushan and Nagappa U. Bhajantri. Multiple Camouflage Breaking by Co-occurrence and Canny.
Results of the Co-occurrence and Canny Method Images from P. Nagabhushan and Nagappa U. Bhajantri. Multiple Camouflage Breaking by Co-occurrence and Canny.
Convexity-based Method • This method uses an operator (Darg) to create an output image whose intensity level is a reflection of the convexity of the original image. • The Darg operator is defined by the sum of Yarg, rotated 0°, 90°, 180°, and 270°. • Yarg is the y-derivative of the polar coordinates of the gradient argument of the original image. Yarg detects the zero-crossing of the gradient argument.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. A model for visual camouflage breaking.
Why Convexity? Thayer’s principle of counter shading (a)A cylinder of constant albedo under top lighting. (b) A counter shaded cylinder under ambient lighting. (c) Thayer’s principle: the combined effect of counter-shading albedo and top lighting breaks up the shadow effect (or convex intensity function). Images from Ariel Tankus and Yehezkel Yeshurun. Convexity-based Camouflage Breaking.
Convexity-based Method • Though edge based methods have their advantages, this method overcomes some of the flaws of an edge-based approach such as, • • Sensitivity to illumination Scale Strong effect of the surroundings Cluttered or textured images
How does the Convexity-based method handle changes in Illumination, Scale or Orientation? Images from Ariel Tankus and Yehezkel Yeshurun. Convexity-based visual Camouflage Breaking.
Convexity-based Method Invariance to derivable strongly monotonically increasing transformation of the gray-level function. Images from Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estimation.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. A Model for Visual Camouflage Breaking.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. A Model for Visual Camouflage Breaking.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estimation.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estimation.
Convexity-based Method Images from Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estimation.
Conclusion • Co-occurrence and Canny Method • Advantages • Simple • Creates a good outline of the object • Disadvantage • Does not extract the object • Must have the known background • Only tested on synthetic images and may not be effective in real application
Conclusion • Convexity-based Method • Advantages: • Robust algorithm • Precise in finding foreground objects • Disadvantage: • Does not extract the object • Threshold must be determined, which can change the results
References [1] [2] [3] [4] [5] [6] P. Nagabhushan and Nagappa U. Bhajantri. Multiple Camouflage Breaking by Cooccurrence and Canny, University of Mysore, Manasa Ganotri, 2004. Ariel Tankus, Yehezel Yeshurun, and N. Intrator. Face Detection by Direct Convexity Estimation, Pattern Recognition Letters 18(9) (1997), 913 -922. Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estimation, IEEE International Workshop on Visual Surveillance, pages 42 -48, Bombay, India, January 1998. In conjunction with ICCV 1998. Ariel Tankus and Yehezkel Yeshurun. A model for visual camouflage breaking, 1 st IEEE International Workshop on Biologically Motivated Computer Vision (BMCV), pages 139 -149, Seoul, Korea, May 2000. Ariel Tankus and Yehezkel Yeshurun. Convexity-based camouflage breaking, International Conference on Pattern Recognition (ICPR), pages 454 -457, Barcelona, Spain, September 2000. Ariel Tankus and Yehezkel Yeshurun. Convexity Based Visual Camouflage Breaking, Computer Vision and Image understanding 82, (2001) 208 -237.
- Slides: 21