National Institute of Technology Rourkela
National Institute of Technology Rourkela
National Institute of Technology Rourkela
0 1 0 1
1 0 1 0
2 2 1 2
3 3 3 2
Describe the various texture classification criteria which are derived from the GLCM.
2. Given an image, I(x, y), an interest point detector typically computes a matrix M(x, y) which captures the
4
changes of pixel intensities in a neighborhood of each pixel location, in the image.
t 3 |t| < 1
f (t )
0 otherwise
What will be the representation of the function in the space V0 using the piece-wise constant representation
scheme (a) taking the average of the function in an interval as it’s piece-wise constant value in that interval (b)
using new scheme (c) Find representation of function in V1 space using new scheme (d) what is mean square
error in both schemes.
b. Give an expression for refinement equation and wavelet equation in multiresolution analysis of a function. 4
Present an analysis of both with their properties in vector space
4.a. Let ( , ) denote a digital image of size 256 × 256. In order to compress this image, we take its Discrete Cosine
Transform ( , ), , = 0, … ,255 and keep only the Discrete Cosine Transform coefficients for , = 0, … , 3
with 0 ≤ < 255. The percentage of total energy of the original image that is preserved in that case is given by
the formula + + 85 with , constants. Furthermore, the energy that is preserved if = 0 is 85%. Find the
constants , .
b. Let ( , ) denote a digital image of size × pixels that is zero outside 0 ≤ ≤ − 1, 0 ≤ ≤ − 1, where 4
and are integers and powers of 2. In implementing the standard Discrete Hadamard Transform of ( , ), we
relate ( , ) to a new × point sequence ( , ).
(i) State the main disadvantage of the Discrete Hadamard Transform.
(ii) In the case of = = 2 and
1 2
f ( x, y ) calculate the Hadamard transform coefficients of ( , ).
2 3
c.
The covariance matrix of the image vector f calculated as part of the K-L Transform is 4
a 0 b2
C f 0 a b2
b2 b2 a
Let us reconstruct the approximations to the three original images by using one or two principal component
images. What would be the mean square error incurred in doing so in each case?
5.a. Use a suitable morphological tool for detecting the number 4 present in the image shown in Fig. 1. Describe the 3
steps for developing the tool.
Fig. 1
b.
4
Fig. 2
The original shape of an object is shown in Fig 2 (A). White represents “1” or “Foreground” and black
represents “0” or “Background.” The original shape is altered, by either a single dilation or a single erosion, to
produce the resulting shapes shown in (B) and (C). For each resulting shape, state whether it was the result of
dilation or erosion of the original shape and give a structuring element that produces the resulting shape. Show
the origin of each structuring element clearly.
6.a. Optimal edge detection techniques for an image follow certain criteria. Discuss them elaborately along with 4
Canny edge detection technique.
b. An image has very non-uniform illumination having two types of multiple objects with a background. Propose a 4
segmentation method based on greyscale thresholding with proper justification.
c. Region merging techniques for image segmentation are highly popular now a days. What are the different 4
methods to determine the region similarity of an image? Briefly describe the region merging of an image using
hypothesis testing.