Image analysis – Image transformation or preprocessing – Changing the image coordinates
Reexamination Certificate
1999-11-08
2004-10-12
Boudreau, Leo (Department: 2621)
Image analysis
Image transformation or preprocessing
Changing the image coordinates
C382S107000
Reexamination Certificate
active
06804419
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to image processing method and apparatus, particularly to method and apparatus for forming a high-resolution image from a low-resolution image, motion vector detecting method and apparatus for use together with the image processing apparatus, method and apparatus for synthesizing a plurality of images, further to a recording medium readable by a computer which is used in these methods and apparatuses, and the like.
2. Related Background Art
Various methods have been heretofore proposed as methods for converting the resolution from inputted low-resolution information to high-resolution information.
In the conventional proposed resolution converting methods, a high resolution is realized by interpolating pixels to the low-resolution information, and the conversion processing method differs with the type of the object image (e.g., a multivalued image in which each pixel has gradation information, a binary image binarized by a pseudo intermediate gradation, a binary image binarized by a fixed threshold value, a character image, and the like).
As the pixel interpolating method in the conventional resolution converting method, a closest interpolating method of arranging the same pixel value closest to an interpolation point as shown in
FIG. 1
, a common primary interpolating method of determining the pixel value of an interpolation point E by the following operation in accordance with the distances of four points (four point pixel values are set to A, B, C, D) surrounding the interpolation point as shown in
FIG. 2
, and the like are generally used.
E=
(1
−i
)(1
−j
)
A+i
(1
−j
)
B+j
(1
−i
)
C+ijD
(1)
(when the distance between pixels is set to 1, the interpolation point E has a distance i in a transverse direction and a distance j in a vertical direction from A (i≦1, j≦1)).
Moreover, as heretofore represented by a sampling theorem, means for converting a sampled discrete signal to a continuous signal comprises passing the signal through an ideal low pass filter which can be represented by SINC function, so that the continuous signal can be reproduced. Moreover, since the operation of SINC function requires much processing time, there is proposed another method which comprises approximating the interpolation function represented by the SINC function, and calculating an interpolated value only by a simple operation of sum of products.
For example, in a known cubic convolution interpolating method, the approximating of the interpolation function can be realized. A method of calculating the interpolated value by the interpolating method will be described with reference to FIG.
3
. In the pixel arrangement shown in
FIG. 3
, P denotes an interpolated point (interpolation point), and P
11
to P
44
denote pixel values of 16 pixels surrounding the point. Then, the interpolation point is interpolated using a cubic convolution function shown in the following equation. Additionally, in the following equation, x{circumflex over ( )}y represents y power of x.
P
=
[
f
(
y1
)
f
(
y2
)
f
(
y3
)
f
(
y4
)
]
[
P11
P12
P13
P14
P21
P22
P23
P24
P31
P32
P33
P34
P41
P42
P43
P44
]
[
f
(
x1
)
f
(
x2
)
f
(
x3
)
f
(
x4
)
]
(
2
)
f
(
t
)
=
sin
(
π
t
)
/
(
π
t
)
≅
[
1
-
2
&LeftBracketingBar;
t
&RightBracketingBar;
^
2
+
&LeftBracketingBar;
t
&RightBracketingBar;
^
3
(
0
≤
&LeftBracketingBar;
t
&RightBracketingBar;
<
1
)
4
-
8
&LeftBracketingBar;
t
&RightBracketingBar;
+
5
&LeftBracketingBar;
t
&RightBracketingBar;
^
2
-
&LeftBracketingBar;
t
&RightBracketingBar;
^
3
(
1
≤
&LeftBracketingBar;
t
&RightBracketingBar;
<
2
)
0
(
2
≤
&LeftBracketingBar;
t
&RightBracketingBar;
)
]
(
3
)
x
1
=1+(
u−[u]
)
y
1
=1+(
v−[v]
)
x
2
=(
u−[u]
)
y
2
=(
v−[v]
)
x
3
=1−(
u−[u]
)
y
3
=1−(
v−[v]
)
x
4
=2−(
u−[u]
)
y
4
=2−(
v−[v]
) (4)
(In the equation, [ ] denotes Gauss' notation, and takes an integer portion.)
However, as a result of resolution conversion by the above-described three types of interpolating methods, a blur by interpolation, and a block-shaped jaggy dependent on input low-resolution image occur, and high quality and resolution information cannot be prepared.
To prepare the high-resolution information from the low-resolution information in such background, there is also proposed an interpolating method including a technique of realizing the resolution conversion without generating the interpolation blur attributed to the interpolating processing or the jaggy, a technique of preparing an excellent edge while maintaining the continuity of pixel values, and the like.
However, the resolution conversion by the above-described conventional interpolating method has the following defect. Specifically, even if the high-resolution information is prepared, the enhancement of image quality is limited.
As apparent from the sampling theorem, since the information with the input resolution equaling or exceeding Nyquist limit does not exist in the input image, the preparation of information with Nyquist frequency or more frequency is all based on presumption. Therefore, it is easy to convert flat artificial images such as not-complicated CG image, illustration image, and animation image to jaggy-less images, but it is difficult to enhance the image quality of a natural image by presuming the information equaling or exceeding the Nyquist limit. Specifically, even if any method is used, the image quality of the image obtained by inputting low-resolution information and converting the resolution to a high resolution is evidently deteriorated as compared with the image inputted originally as the high-resolution information.
On the other hand, with the spread of digital video cameras in recent years, it becomes easy to input the picked-up motion image into a computer in the unit of continuous one frame. Therefore, one frame of motion image can also be outputted via a printer. However, as compared with the yearly increasing output resolution of the printer, the input resolution of a picking up system tends to increase, but it is still low in the present situation.
Therefore, as described above in the conventional example, instead of preparing one frame of high-resolution still image from one frame of low-resolution still image, it is considered that one frame of high-resolution still image is prepared from a plurality of continuous low-resolution still images taken from the motion image.
The technique of preparing the high-resolution still image from the low-resolution motion image is proposed in Japanese Patent Application Laid-Open No. 05-260264. The proposed method comprises comparing images continuous in point of time, detecting parameters of affine transformation and parallel movement based on the difference of the images, and synthesizing these images. Additionally, an example in which the synthesizing method is utilized for interpolation is also mentioned.
However, this proposal has the following problem:
Specifically, in the method of utilizing the synthesizing method for the interpolation, by comparing the continuous images enlarged by the interpolating method shown in
FIGS. 1
to
3
, the parameters are calculated to determine an interpolation position, before performing the synthesis. However, for the enlarged image obtained by the interpolation in this manner, new high-resolution information is not prepared by the interpolating operation itself. Therefore, even when the synthesis processing is performed using the enlarged image in this manner, a really high-resolution image is not necessarily obtained.
Here, the interpolat
Boudreau Leo
Canon Kabushiki Kaisha
Fitzpatrick ,Cella, Harper & Scinto
Lu Tom Y.
LandOfFree
Image processing method and apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3286674