Transcript Document
Introduction
2007
Theo Schouten
1
Human Eye
•6-7 106 Cones, 120 106 Rods
•Reaction time: 0.1 sec (enough for transferring 100 nerve stimuli)
•Parallel processing by 1010 nerve cells (neurons)
•Adjustment to light intensity (factor 1010)
•Stereo: 2 eyes and their continuous movement, focusing
•Continuous
assimilation of Theo
moving
images
2007
Schouten
2
Visual effects
2007
Theo Schouten
3
Steps in image processing
•
•
•
•
•
•
Production (CCD camera, CT, MRI )
Storage, transmission and compression
Transformation eg to a Fourier or Wavelet space
Restoration, e.g. lens distortions or movement
Enhancement: making suitable for an application
Segmentation: finding (parts of) relevent objects
– finding edges: local deviations in an image
– finding surface areas: local similarities
• Describing and measuring of regions and objects
• Image understanding (Pattern recognition, classical,
neural networks, AI)
2007
Theo Schouten
4
Special methods
• 2 and 3 D Computed Tomography (CT) images to
make it possible to view the interior of an object.
• binary images, e.g. mathematical morphology
• color and in general multispectral images
• texture (fine structures on a surface)
• 3-D reconstruction out of stereo images
• video processing (movement, 3-D)
• 3D image processing
2007
Theo Schouten
5
Computing challenges
• the large amount of data
• high complexity of algorithms
• high speed requirements
– parallel machines, algorithms
– special purpose chips, CPU’s
• real-time requirements
– image recorders, robots, production machines
2007
Theo Schouten
6
Optimizing code
/* slow.c */
#define SIZE 4096
float imin[SIZE][SIZE], imout[SIZE][SIZE];
int main(int argc, char **argv)
{
int i, j, k, l;
/* initialize image */
for(i=0; i < SIZE; i++) for(j=0; j < SIZE; j++) imin[j][i] = (i+j) % 256;
/* average each pixel with its neighbours */
for(i=0; i < SIZE; i++) {
for(j=0; j < SIZE; j++) {
if( i == 0 || j == 0 ) imout[j][i] = imin[j][i];
else if( i == SIZE-1 || j == SIZE-1 ) imout[j][i] = imin[j][i];
else {
imout[j][i] = 0;
for(k=-1; k < 2 ; k++) for(l=-1; l < 2; l++) imout[j][i] += imin[j+l][i+k];
imout[j][i] /= 9 ;
}}}}
2007
Theo Schouten
7
Relation other science fields
•
•
•
•
•
•
•
•
•
•
•
Signal Analysis; 1D signal.
Geometry; position, orientation and size of objects from 3-D measurements
Linear Algebra
Estimation Theory; eg compression of images or in determining the
movement of objects.
Statistical Pattern recognition; classification
Syntactical Pattern recognition; structure of an image
Artificial Intelligence; representation and manipulation of knowledge
information that is extracted from images.
Real Time Systems; interaction with other machines.
System and Computer Architecture; parallel (1 CPU per pixel); special
processors
Computer Graphics; for projecting the images and showing the results of
the manipulations
User Interfaces; for good interaction with the software systems used for
image manipulations.
2007
Theo Schouten
8
Uses of image processing
• Medical Imaging
• Machine en Robot vision (industrial production and
inspection )
• Remote Sensing (satellite and aerial photos)
• graphical and game industry, image editing in
combination with computer graphics
• consumer photo and video editing
2007
Theo Schouten
9
Integrated Examination
•
•
•
•
•
4 assignments
1st individual, rest in groups
working towards a goal
all assignments must have been handed in
final grade is average if all grades are 5 or more
– otherwise the lowest grade
2007
Theo Schouten
10
FEED
Fast Exact Euclidean Distance transformation
Joint research with Egon van den Broek en Harco Kuppens
2007
Theo Schouten
11
Video Surveillance
2007
Theo Schouten
12
Depth from floor contact
2007
Theo Schouten
13