Picture Similarity

0
0
1889 days ago, 474 views
PowerPoint PPT Presentation

Presentation Transcript

Slide 1

Picture Similarity Longin Jan Latecki CIS Dept. Sanctuary Univ., Philadelphia latecki@temple.edu

Slide 2

Image Similarity Image based, e.g., distinction of benefits of comparing pixels Histogram construct Based with respect to closeness of items contained in pictures, requires picture division

Slide 3

Mathematical Representation of Images A picture is a 2D flag (light power) and can be spoken to as a capacity f ( x, y). c oordinates ( x, y) speak to the spatial area of p oint ( x, y) that is called pixel ( pi cture e lement) estimation of f ( x, y) is the light power called dark esteem (or dim leve l) of picture f · Images are of two sorts: constant and discrete A ceaseless picture is a component of two factors, that take values in a continuum. E .g. : The power of a photographic picture recorded on a film is a 2D work f ( x, y) of two genuine esteemed factors x and y.

Slide 4

· A discrete picture is an element of two factors, that take values over a discrete set (a number lattice) E .g. : The power of a discretized 320 x 240 photographic picture is 2D work f ( i , j ) of two number esteemed factors i and j . In this manner, f can be spoken to as a 2D lattice I [ 320,240 ] A shading picture is normally spoken to with three grids: Red[ 320,240 ], Green[ 320,240 ], Blue[ 320,240 ]

Slide 5

Pixel based picture closeness Let f and g be two dark esteem picture capacities.

Slide 6

Let an and b wager two pictures of size w x h . Give c a chance to be some picture attributes that doles out a number to every picture pixels, e.g., c(a,x,y ) is the dim estimation of the pixel. Pixel to pixel contrasts:

Slide 7

We can utilize factual mean and fluctuation to add solidness to pixel to pixel picture distinction:

Slide 8

Let v(a) be a vector of all c(a,x,y) values relegated to all pixels in the picture a . Picture similitude can be communicated as standardized inward results of such vectors. Since it yields greatest qualities for equivalent casings, a conceivable difference measure is

Slide 9

Image histogram is a vector If f :[1, n]x[1, m] ��  [0, 255] is a dim esteem picture, then H ( f ) : [0, 255] ��  [0, n*m] is its histogram, where H ( f ) (k) is the quantity of pixels (i, j) to such an extent that F(i, j)=k Similar pictures have comparative histograms Warning: Different pictures can have comparable histograms

Slide 10

Image Histogram (3, 8, 5)

Slide 11

Hg Hr Hb

Slide 12

Histograms Histogram Processing 1 4 5 0 3 1 5 1 Number of Pixels dim level

Slide 13

Histogram-based picture likeness Let c be some picture attributes and h(a) its histogram for picture a with k histogram receptacles.

Slide 14

Homework 5 Implement in Matlab a basic picture internet searcher (no GUI required). Just look at the execution of no less than two picture separates on a little arrangement of pictures.

SPONSORS