Image Transformation Performed Using Spatial Transformations

1
Image Transformation Performed Using Spatial
Transformations

• ECE 552
• Project Proposal
• April 24, 2007
• Team RED
• Sam Caylor
• Roopa Channappa
• Vicente Mahoney
• Christina Ward

2
Spatial Transformations

• Spatial Transformations enables the coordinates
  of pixels in one image to be mapped to form a new
  image
• They can perform the following operations on
  images rotation, scaling, and translation

3
Proposal

• Implement Spatial Transformation techniques on
  an image.
• Linear Conformal Transformation
• Affine Transformation
• Projective Transformation
• Polynomial Transformation
• Simulate the techniques using MATLAB on the
  image first.
• Implement the techniques in VHDL
• Compare and contrast the results.

4
Assigned Tasks

• Obtain Matlab Spatial Transformation Results
  Roopa
• Linear Conformal Transformation Roopa
• Affine Transformation - Christina
• Memory Bank - Vicente

5
Spatial Transformation Results Using Matlab

• Original Image

6
Linear Conformal Transformation

• Matlab Code that was created to produce the
  following image.
• I imread(eiffeltower.jpg')
• imshow(I)
• title('original')
• scale 1.2 scale factor
• angle 40pi/180 rotation angle
• tx 1 x translation
• ty 1 y translation
• sc scalecos(angle)
• ss scalesin(angle)
• T sc -ss
• ss sc
• tx ty
• t_lc maketform('affine',T)
• I_linear_conformal imtransform(I,t_lc,'FillValue
  s',.3)
• figure, imshow(I_linear_conformal)

7
Affine Transformation

• I imread(eiffeltower.jpg')
• imshow(I)
• title('original')
• scale 1.2 scale factor
• angle 40pi/180 rotation angle
• tx 0 x translation
• ty 0 y translation
• T 1 0.5
• 0.5 1
• 0 0
• t_aff maketform('affine',T)
• I_affine imtransform(I,t_aff,'FillValues',.3)
• figure, imshow(I_affine)
• title('affine')

8
Projective Transformation

• I imread('eiffeltower.jpg')
• imshow(I)
• title('original')
• T 1 0 0
• 0.5 1 0
• 0 0 0.5
• t_proj maketform('projective',T)
• I_projective imtransform(I,t_proj,'FillValues',.
  3)
• figure, imshow(I_projective)
• title('projective')

9
Polynomial Transformation

• I imread('C\Documents and Settings\Roopa
  Channappa\Desktop\lena.pgm')
• imshow(I)
• title('original')
• T 0 0
• 1 0
• 0 1
• 0.015625 0
• 0.015625 0
• 0.015625 0
• t_poly.tdata T
• I_polynomial imtransform(I,t_poly,'FillValues',.
  3)
• figure, imshow(I_polynomial)
• title('polynomial')

10
Specifications
11
Our Approach

• Spatial Transformations require no operation on
  the image but rather changing the location of the
  image to a new address.
• To satisfy this condition we changed the size of
  memory bank from 4kx32 to 16kx8 and then
  performed our address operation in the
  Logic_function.vhd

12
Block Diagram using FPGA Adv
13
Pre-Synthesis Results

• Modified Memory Bank

14
Pre-Synthesis Cont.

• Logic Function

15
Pre-Synthesis for New Approach
16
Pre-Synthesis Coverage Report

• ADD COVERAGE REPORT HERE

17
Conclusions

• First Approach
• Modified Memory Output bank to be different in
  structure than Memory Input bank
• Modified Logic Function to accept address,
  reassign address as x and y coordinates, perform
  an operation on the coordinates, and then map to
  an individual address
• This parallelism was intensive and forced us to
  proceed with another approach
• Second Approach
• Went back to the same Memory Output bank as
  Memory Input bank but just made it larger in size
• Modified Logic Function to accept address but to
  only operate on one pixel location at a time
• The resulting image was not what was expected