1 MSU CSE 803 Fall 2014 Vectors [and more on masks] Vector space theory applies directly to several image processing/representation problems
2 MSU CSE 803 Fall 2014 Image as a sum of “basic images” What if every person’s portrait photo could be expressed as a sum of 20 special images? We would only need 20 numbers to model any photo sparse rep on our Smart card.
3 MSU CSE 803 Fall 2014 Efaces 100 x 100 images of faces are approximated by a subspace of only x 100 “images”, the mean image plus a linear combination of the 3 most important “eigenimages”
4 MSU CSE 803 Fall 2014 The image as an expansion
5 MSU CSE 803 Fall 2014 Different bases, different properties revealed
6 MSU CSE 803 Fall 2014 Fundamental expansion
7 MSU CSE 803 Fall 2014 Basis gives structural parts
8 MSU CSE 803 Fall 2014 Vector space review, part 1
9 MSU CSE 803 Fall 2014 Vector space review, Part 2 2
10 MSU CSE 803 Fall 2014 A space of images in a vector space M x N image of real intensity values has dimension D = M x N Can concatenate all M rows to interpret an image as a D dimensional 1D vector The vector space properties apply The 2D structure of the image is NOT lost
11 MSU CSE 803 Fall 2014 Orthonormal basis vectors help
12 MSU CSE 803 Fall 2014 Represent S = [10, 15, 20]
13 MSU CSE 803 Fall 2014 Projection of vector U onto V
14 MSU CSE 803 Fall 2014 Normalized dot product Can now think about the angle between two signals, two faces, two text documents, …
15 MSU CSE 803 Fall 2014 Every 2x2 neighborhood has some constant, some edge, and some line component Confirm that basis vectors are orthonormal
16 MSU CSE 803 Fall 2014 Roberts basis cont. If a neighborhood N has large dot product with a basis vector (image), then N is similar to that basis image.
17 MSU CSE 803 Fall 2014 Standard 3x3 image basis Structureless and relatively useless!
18 MSU CSE 803 Fall 2014 Frie-Chen basis Confirm that bases vectors are orthonormal
19 MSU CSE 803 Fall 2014 Structure from Frie-Chen expansion Expand N using Frie- Chen basis
20 MSU CSE 803 Fall 2014 Sinusoids provide a good basis
21 MSU CSE 803 Fall 2014 Sinusoids also model well in images
22 MSU CSE 803 Fall 2014 Operations using the Fourier basis
23 MSU CSE 803 Fall 2014 A few properties of 1D sinusoids They are orthogonal Are they orthonormal?
24 MSU CSE 803 Fall 2014 F(x,y) as a sum of sinusoids
25 MSU CSE 803 Fall 2014 Spatial direction and frequency in 2D
26 MSU CSE 803 Fall 2014 Continuous 2D Fourier Transform To compute F(u,v) we do a dot product of our image f(x,y) with a specific sinusoid with frequencies u and v
27 MSU CSE 803 Fall 2014 Power spectrum from FT
28 MSU CSE 803 Fall 2014 Examples from images Done with HIPS in 1997
29 MSU CSE 803 Fall 2014 Descriptions of former spectra
30 MSU CSE 803 Fall 2014 Discrete Fourier Transform Do N x N dot products and determine where the energy is. High energy in parameters u and v means original image has similarity to those sinusoids.
31 MSU CSE 803 Fall 2014 Bandpass filtering
32 MSU CSE 803 Fall 2014 Convolution of two functions in the spatial domain is equivalent to pointwise multiplication in the frequency domain
33 MSU CSE 803 Fall 2014 LOG or DOG filter Laplacian of Gaussian Approx Difference of Gaussians
34 MSU CSE 803 Fall 2014 LOG filter properties
35 MSU CSE 803 Fall 2014 Mathematical model
36 MSU CSE 803 Fall D model; rotate to create 2D model
37 MSU CSE 803 Fall D Gaussian and 1 st derivative
38 MSU CSE 803 Fall nd derivative; then all 3 curves
39 MSU CSE 803 Fall 2014 Laplacian of Gaussian as 3x3
40 MSU CSE 803 Fall 2014 G(x,y): Mexican hat filter
41 MSU CSE 803 Fall 2014 Convolving LOG with region boundary creates a zero-crossing Mask h(x,y) Input f(x,y)Output f(x,y) * h(x,y)
42 MSU CSE 803 Fall 2014
43 MSU CSE 803 Fall 2014 LOG relates to animal vision
44 MSU CSE 803 Fall D EX. Artificial Neural Network (ANN) for computing g(x) = f(x) * h(x) level 1 cells feed 3 level 2 cells level 2 cells integrate 3 level 1 input cells using weights [-1,2,-1]
45 MSU CSE 803 Fall 2014 Experience the Mach band effect
46 MSU CSE 803 Fall 2014 Simple model of a neuron
47 MSU CSE 803 Fall 2014 Output conditioning: threshold versus smoother output signal
48 MSU CSE 803 Fall D situation in the eye Neuron c has + input to neuron A but - input to neuron B. Neuron d has + input to neuron B but – input to neuron A. Neuron b gives no input to neuron B: it is not in the receptive field of B.
49 MSU CSE 803 Fall 2014 Receptive fields
50 MSU CSE 803 Fall 2014 Experiments with cats/monkeys Stabilize/drug animal to stare Place delicate probe in visual network Move step edge across FOV Probe shows response function when the edge images to receptive field Slightly moving the probe produces similar signal when edge is nearby
51 MSU CSE 803 Fall 2014 Canny edge detector uses LOG filter
52 Cornsweet Illusion See, e.g., wikipedia
53 MSU CSE 803 Fall 2014 Summary of LOG filter Convenient filter shape Boundaries detected as 0-crossings Psychophysical evidence that animal visual systems might work this way (your testimony) Physiological evidence that real NNs work as the ANNs
54 MSU CSE 803 Fall 2014 Morphology Operations
55 Morphology Operations
56 MSU CSE 803 Fall 2014 Applications
57 Applications