FIELD OF INVENTION
The present invention relates to a linear, method and mechanism for performing
perspective transformation of image for low-end embedded systems.
BACKGROUND OF THE INVENTION
The term perspective means the appearance of things relative to one another as
determined by their distance from the viewer. An embedded device comprises a
mechanism for selecting an image of interest and a mechanism for applying
perspective transformation functions to the image of interest. The perspective
transformation functions are carried out such that the image of interest varies
from a greater width at a foreground viewpoint of the image of interest to a
lesser width toward a vanishing point of the image of interest and such that the
image of interest varies from a greater amount of expansion and lesser amount
of compression at the foreground viewpoint of the image of interest to a lesser
amount of expansion and a greater amount of compression towards the
vanishing point of the image of interest.
Document PCT/US 2005 / 035 285 discloses a method for perspective
transformation of two-dimensional images.

In this document the determination of new x and y coordinates involves
nonlinear calculation, which is very complex and involves non-integer
calculations. Even if it is possible to perform on embedded devices, the
performance will not be good.
US Patent 5, 808, 623 provides a system and method for perspective transform
in a computer using multipass algorithm.
This document does not mention about transforming a selected portion of the
image.
A third dimension pop-up generation from a two-dimensional transformed image
display is disclosed in US Patent 4, 841, 292.
This document does not describe any way to do scaling and perspective
transformation. This means, if there is a large image (more than the screen size
of the embedded device) and perspective transformation is required to be
performed then it has to be scaled down to the screen size and then perspective
transform is to be performed.

Most image transformations can be described as sets of mathematical
transformation functions represented by sets of mathematical equations; these
equations are descriptive of the operations being performed on the image
regardless of the particular platform on which the transformations are
implemented. The mathematical equations describing one example set of
perspective transformation functions for the illustrated embodiment are given
below as Equations (1) and (2). For each pixel in image:


In Equations (1) through (6), (x IITT, y in) is the input pixel location, (x oul, y oul)
is the output pixel location, and x shifl is the x-coordinate of the left endpoint of
the top width of the one-point perspective image. W is the width of image,
which is also the bottom width of image. H is the height of both image and
image. D is the length of the smaller of the two parallel edges in image. Since
the vanishing point is above the image, the smaller parallel edge is the top edge
in this case. When the vanishing point is to the left, right, or bottom of the
original image, the resulting perspective image will taper towards the left, right,
or bottom. In such cases, the smaller parallel edge will be on the left, right or
bottom of the image.
SUMMARY OF THE INVENTION
The present invention overcomes the above mentioned problems in prior art.
Perspective transformation is the method to draw a 3D object on a 2D surface.
The main object of the present invention is to provide a linear; singlepass and
efficient method to perform perspective transformation on the complete or

selected portion of an image object. The core of this method lies in performing
progressive perspective transform on each of the quadrilateral images. In
addition, the effect can be applied to a selected portion of the image as well as
to perform scaling and perspective transformation simultaneously.
The perspective transform of the image is achieved by a signlepass procedure by
means of a single phase of a one-dimensional scaling of each dimension. Each
step consists of scaling a single vertical line of the image in horizontal direction
(x-dimension) first and then performs scaling in vertical direction. The vertical
scaling is progressive, means the amount of scaling of each line changes with x-
direction. This leads to a trapezoidal shape of the image achieving the
perspective transform.
This method can be applied to two images together and achieve 3D cube
rotation effect in horizontal or vertical direction. This method can be applied on
a single image to obtain flipping effect. It is also applicable to a selected portion
of the image, this mean that a single object inside the complete image can also
be separately applied this effect without modifying the complete image.

In a preferred embodiment the present invention relates to a linear method for
performing perspective transformation of image in a low-end embedded system,
comprising the steps of; applying horizontal scaling on a single vertical line;
applying vertical scaling on the same line; and varying the amount of vertical
scaling as the distance of the vertical line varies from the side of a quadrilateral
image.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
The invention can now be described in detail with the help of the figures of the
accompanying drawings in which
Figure 1 is a block diagram of an embedded device of the present invention.
Figure 2 shows the image shapes that can be
achieved by the present invention,
Figure 3 shows perspective transformation applied
on a selected portion of an image.
Figure 4 transformation and scaling achieved in a
singiepass of the present invention.

Figure 5 illustrates 3D quadrilateral rotation
achieved using the present invention.
Figure 6 shows 3D quadrilateral rotation performed on two images,
Figure 7 shows in flow chart form perspective
transformation of the present invention.
Figure 8 illustrates in detail the perspective transformation method.
DETAILED DESCRIPTION
As shown in Figure 1 the embedded system of the present invention comprises a
microprocessor 1, system main memory 2, a system bus 3, and a graphics
controller unit 4 with an image buffer 5 and a control logic 6. The
microprocessor 1 is connected to the system main memory 2 via the system bus
3. The microprocessor 1 is also connected to the image buffer 5 in the graphics
controller entity 4.

The actual implementation of the perspective transformation method of the
present Invention is shown in Figure 8 and can be explained as follows.
Initialization part of Method:
X, y are position of the pixel in source image
Xout, Yout are pixel position of the transformed image in the destination image
hr, hi left height and right height of the out polygon (in pixels)
wt, wb are top and bottom widths of out polygon
ILOOP AND JLOOP are I and J Direction dimensions of output surface
inSurface_ILOOP and inSurface_JLOOP are I and J Direction dimensions of
input Surface.
1. Calculate the left and right heights, top and bottom width (hr, hl, wt,
and wb)
2. Initializing ILLOOP; Check whether either heights or widths are equal
to determine the ILOOP and set the inSurfaceJLOOP and
inSurfaceJLOOP.
3. Initializing JLOOP: if the ILOOP is set in width direction (i.e. x
direction) then 3LOOP will be the max value of hr, hi Else 3LOOP will be
the max of wt, wb.
4. Calculate x position in inSurface by scaling (I+InRect->x) with
inSurfaceJLOOP / ILOOP ratio.
5. Calculate Line length in J-ditrection(JLINE) by subtracting line
difference from 3LOOP

6. Calculate y position in inSurface by scaling (J+InRect->y) with
inSurfaceJIOOP / JLINE ratio.
7. Copy pixel from x,y of inSurface to X, Y of outBuf
8. Do step4 to Step7 for ail pixel in the inSurface.
Some of the advantageous of the present invention are enumerated below:
Because of the use of linear calculation in the present invention, the performance
of perspective transformation can be improved by two to three times over the
perspective transformation of prior art method.
In the present invention the perspective transformation can be applied to certain
portion or certain objects in the image. The sharing of image to rhombus or
parellogram shapes can be done with order of pixels time. The scaling and
perspective transformation are achieved in a single pass.
The method is highly customized for regular shapes especially meant for cube
rotation effect. It will not perform all types of perspective transformation, i.e. it
will not work for irregular shapes.

The linear method for calculation in the present invention is governed by the
following equations:
Xout = (Win / Wout) * Xin+ Xstart
Yout= (Hin / Hout - Hdiff * I)) 4- Ystart
The scaling and perspective transformations are achieved in a singlepass in the
method algorithm of the present invention. The perspective transformation can
be achieved on a selected area or object of interest.

WE CLAIM
1. A linear method for performing perspective transformation of image in a
low-end embedded system, comprising the steps of:
- applying horizontal scaling on a single vertical line;
- applying vertical scaling on the same line; and
- varying the amount of vertical scaling as the distance of the vertical
line varies from the side of a quadrilateral image,
2. The method as claimed in claim 1, wherein said linear method for
determination of X, Y is governed by
Xout = (Win / Wout) * Xin+ Xstart
Yout = (Hin / Hout - Hdiff * i)) + Ystart
3. The method as claimed in claim 1, wherein the scaling and perspective
transformations are achieved in a single pass of the method algorithm.

4. The method as claimed in claim 1, wherein said perspective
transformation can be achieved on a selected area or object of interest.
5. A low-end embedded system for performing perspective transformation of
image, said system comprising:

- a microprocessor for performing perspective transformation of an
image by applying horizontal and vertical scaling on a single vertical
line;
- a system main memory connected to said micro processor through
a system bus; and
- a graphics controller unit having an image buffer connected to said
microprocessor through said system bus for displaying perspective
transformation of the image in a graphics screen.

6. A linear method of performing perspective transformation of image for
lowend embedded system, substantially as herein described and
illustrated in the figures of the accompanying drawings.
Dated this 17th day of October 2008

The present invention overcomes the above mentioned problems in prior art. Perspective transformation is the method to draw a 3D object on a 2D surface. The main object of the present invention is to provide a linear; singlepass and
efficient method to perform perspective transformation on the complete or selected portion of an image object. The core of this method lies in performing progressive perspective transform on each of the quadrilateral images. In addition, the effect can be applied to a selected portion of the image as well as to perform scaling and perspective transformation simultaneously.
The perspective transform of the image is achieved by a signlepass procedure by means of a single phase of a one-dimensional scaling of each dimension. Each step consists of scaling a single vertical line of the image in horizontal direction (x-dimension) first and then performs scaling in vertical direction. The vertical
scaling is progressive, means the amount of scaling of each line changes with x-direction. This leads to a trapezoidal shape of the image achieving the perspective transform. This method can be applied to two images together and achieve 3D cube rotation effect in horizontal or vertical direction. This method can be applied on
a single image to obtain flipping effect. It is also applicable to a selected portion of the image, this mean that a single object inside the complete image can also be separately applied this effect without modifying the complete image.
In a preferred embodiment the present invention relates to a linear method for performing perspective transformation of image in a low-end embedded system, comprising the steps of; applying horizontal scaling on a single vertical line; applying vertical scaling on the same line; and varying the amount of vertical scaling as the distance of the vertical line varies from the side of a quadrilateral image.