Download - image mosaics
-
7/29/2019 image mosaics
1/27
Image Mosaic
Snehith Reddy
Aditi Singh, A-36
Samridhi Sharma,A-50
Rahul Menon,B-63
-
7/29/2019 image mosaics
2/27
-
7/29/2019 image mosaics
3/27
We describe a process for creating an image mosaic as a
collection of small images arranged in such a way that
when they are seen together from a distance they suggest alarger image.
To visually suggest the larger form, the small images are
arranged to match a large picture as much as possible, andthen their colors are adjusted to better suggest the overall
form.
Arrangement of the small images may be either manual or
automatic.
Adjustment of the colors in the small image to further
suggest the larger picture is fully automatic
-
7/29/2019 image mosaics
4/27
-
7/29/2019 image mosaics
5/27
-
7/29/2019 image mosaics
6/27
Image mosaics uses layered imagery, where the subject of
the work is both the tiny features that only can be seen up
close and the large scale features that only can be seen at a
distance.
A picture (usually a photograph) that has been divided into
(usually equal sized) rectangular sections, each of which isreplaced with another image that matches the target photo.
When viewed at low magnifications, the
individual pixels appear as the primary image, while close
examination reveals that the image is in fact made up of
many hundreds or thousands of smaller images.
-
7/29/2019 image mosaics
7/27
There are two kinds of mosaic, depending on how the matching
is done.
In the simpler kind,
Each part of the target image is averaged down to a single
color.
Each of the library images is also reduced to a single color.
Each part of the target image is then replaced with one from
the library where these colors are as similar as possible.
In effect, the target image is reduced in resolution
(by downsampling), and then each of the resulting pixels is
replaced with an image whose average color matches that
pixel.
-
7/29/2019 image mosaics
8/27
In the more advanced kind of photographic mosaic, the target
image is not downsampled,
the matching is done by comparing each pixel in the rectangle
to the corresponding pixel from each library image.
The rectangle in the target is then replaced with the library
image that minimizes the total difference.
This requires much more computation than the simple kind,
but the results can be much better since the pixel-by-pixel
matching can preserve the resolution of the target image.
-
7/29/2019 image mosaics
9/27
We address the problem of forming an image mosaic:-
Given a collection of tile images Ti and a target imageI, createan image mosaicMthat resemblesIat a coarse scale but is
composed of a tiling of the Tis. We solve the problem in four
steps:
(1) choose images
(2) choose a tiling grid,
(3) find an arrangement for the image tiles within the grid, and
(4) correct the tiles to match the target image.
-
7/29/2019 image mosaics
10/27
Some images simply work better than others.
Recognition plays an important role, so iconic figures
such as political leaders, actors, famous works of art, and
well-known scenes are desirable.
Furthermore, figures that are easy to recognize at very
low resolution tend to work well.
Choosing images
-
7/29/2019 image mosaics
11/27
Finding a tiling pattern:-
Generally we use semiregular rectangular grids
A more general (nonperiodic) tiling might be found that
achieves a better match for a given target image
Arranged in a rectangular grid, all the small image tiles
must have the same aspect ratio.
Thus, it is important to choose an aspect ratio appropriatefor the entire set of image tiles, as they will either have to
be cropped or stretched to conform to this ratio.
-
7/29/2019 image mosaics
12/27
Among rectangular tilings, there are two varieties:
A regular grid, and
Angled grids
Studies in perception show that the eye is least sensitive to
grids angled at 45o and most sensitive to horizontal and
vertical grids.
Thus, traditional screening methods have typically oriented
the screen at an angle to reduce its visual impact.
-
7/29/2019 image mosaics
13/27
-
7/29/2019 image mosaics
14/27
Arranging the image tiles
Having selected a tiling grid, we need to place individual image
tiles into the grid.
There are a number of arrangement options:
Use the same tile everywhere
Choose a random arrangement of the image tiles.
Arrange different tiles manually by eye
Place image tiles by matching their average colors to the region
of the target image that they cover Find a more detailed match between the tiles and the target
image based on the shapes and colors within the images
-
7/29/2019 image mosaics
15/27
Use the same tile everywhere
-
7/29/2019 image mosaics
16/27
Choose a random arrangement of
the image tiles.
-
7/29/2019 image mosaics
17/27
Arrange different tiles manually by eye
-
7/29/2019 image mosaics
18/27
Place image tiles by matching their average colors to
the region of the target image that they cover.
-
7/29/2019 image mosaics
19/27
Find a more detailed match between the tiles and the target
image based on the shapes and colors within the images
-
7/29/2019 image mosaics
20/27
Our objective is to match the color (brightness, in the case of
our grayscale image) of the tile image to the color of the regionin the target image that is covered by the tile.
If the target image has a constant colorx across this region,
then we want to adjust the color of the tile image so that its
average color isx.
If the brightness of the target image ranges from dark on the left
side of the region to light on the right side of the region, then we
would like the brightness of the image tile to somehow matchthis gradient as well.
In addition to coarsely matching the colors of the target
image, we would like to preserve as much as possible the
features of the tile images.
Color Correction
-
7/29/2019 image mosaics
21/27
We make use of a correction rule, which takes as input an
image tile and a desired average color a, and generates a
correction function that maps a colorx in the image tile to
a color F(x) in the final mosaic such that the region of the
mosaic covered by the image tile will have the average
color a.
There are a variety of families of correction functions thatcould fulfill this role.
For example, the constant function F(x)=a would achieve
the correct overall average, but ignores the original colorsof the image tiles.
-
7/29/2019 image mosaics
22/27
A more sensible solution might be to scale the colors in the
input tile so that the desired average is achieved.
Suppose, for example, that the desired average color a is less
than the average color atof the image tile.
Then our scaling correction rule would generate the
correction function F(x)=(a/at)x.
A different correction ruleshiftingyields a correction
function that shifts all the colors in the tile image as follows:
F(x)=x+(a-at).
-
7/29/2019 image mosaics
23/27
It is found that a combination of the shifting and
scaling rules is sufficient for our purposes and is easy to
compute.
So, let us return to the example where the desired average
color a is less than the average color atof the image tile.
The shift-and-scale rule works as follows: if theminimum color mtof the image tile is greater than at-a,
then we use the shift rule above: F(x)=x+(a-at);
Otherwise we use a combination of shifting and scaling:F(x)=a(x-mt)/(at-mt).
-
7/29/2019 image mosaics
24/27
Correction rule is simply applied for every pixel in every tile
in the mosaic, supplying as the desired average color a the
color of the corresponding pixel in the target image.
In regions where the target image has a fairly constant color,
the tile image will simply be shifted and scaled to achieve the
target color.
In regions where more complex shade variations occur, the tile
image will vary similarly.
If the tile images are larger than 16x16 pixels, then an
efficiency improvement may be implemented by building a
table once for each tile image.
-
7/29/2019 image mosaics
25/27
The table contains the two parameters of the correction
function-how much to shift, and how much to scalefor every
possible desired average a.
Then, for every pixelp in the tile image, we use the
corresponding pixel in the target image as our index a into the
table, and then shift and scalep according to the entries in the
table.
-
7/29/2019 image mosaics
26/27
Encoding extra information in an image for
transmission
Encoding extra information in an image for security,
New forms of halftoning screens for printing,
Association of images for advertising.
USES:-
-
7/29/2019 image mosaics
27/27