Resolution-converting apparatus and method

1. A resolution-converting method, comprising:

applying an edge-directed interpolation to an input image and producing an intermediate image; and

converting a sampling rate with respect to the intermediate image and producing an output image having a predetermined resolution,

wherein the edge-directed interpolation comprises,

a first interpolation of generating interpolated pixel of a part of the input image, and

a second interpolation of generating interpolated pixels of other remaining part of the input image.

2. The resolution-converting method as claimed in claim 1 , wherein the edge-directed interpolation comprises:

using absolute differences of pixels of the input image to estimate an edge direction;

adaptively determining an interpolation coefficient based on a pixel value of a position corresponding to the estimated edge direction;

using the edge direction and the interpolation coefficient and calculating an edge-directed interpolation value; and

adaptively blending the edge-directed interpolation value and an interpolation value by a predetermined linear interpolation method and calculating a final interpolation value.

3. The resolution-converting method as claimed in claim 2 , wherein the edge direction is estimated with respect to two diagonal directions and four non-diagonal directions.

4. The resolution-converting method as claimed in claim 2 , wherein the predetermined linear interpolation method is a bi-cubic interpolation method.

5. The resolution-converting method as claimed in claim 2 , further comprising determining whether an interpolation region has a high spatial frequency, and, if the region has a high spatial frequency, using an interpolation value by the predetermined linear interpolation method to calculate a final interpolation value.

6. The resolution-converting method as claimed in claim 5 , wherein it is determined whether the region has the high spatial frequency by using mean-deviated local variance in the interpolation region.

7. A resolution-converting method, comprising:

applying an edge-directed interpolation to an input image and producing an intermediate image; and

converting a sampling rate with respect to the intermediate image and producing an output image having a predetermined resolution,

wherein the calculating the output image comprises:

multiplying a ratio of up-samplings and down-samplings calculated according to the resolution of the input image and a predetermined resolution and calculating the number of filter tabs;

multiplying a window function by a sinc function and calculating a first-order filter coefficient by the number of filter tabs;

using a value obtained from multiplying a Gaussian function and the window function in the first-order filter coefficient and calculating a final filter coefficient; and

converting a sampling rate of the intermediate image, performing filtering in horizontal and vertical directions according to the final filter coefficient, and calculating an output image of a predetermined resolution.

8. The resolution-converting method as claimed in claim 7 , wherein the number of filter tabs is calculated based on an equation as below:

L =round(Max { D,U }×SmoothingAmount× n Lobes−1)×2−1,

where L indicates a filter length, nLobes the number of side lobes, U and D ratios of optimized up-samplings to down-samplings, and SmoothingAmount a constant changing a cut-off frequency of the filter.

9. The resolution-converting method as claimed in claim 7 , wherein the first-order filter coefficient is calculated based on an equation as below:

h

⁡

[

i

]

=

{

sin

⁡

(

x

)

x

}

×

kaiser

⁡

(

1

,

β

)

,

i

=

0

,

1

,

…

⁢

,

L

-

1

,

⁢

x

=

i

-

L

-

1

2

L

-

1

2

×

π

×

Lobes

where

sin

⁡

(

x

)

x

indicates an ideal low-frequency bandpass function, and kaiser(1,β) is a Kaiser window function.

10. A resolution-converting method, comprising:

applying an edge-directed interpolation to an input image and producing an intermediate image;

converting a sampling rate with respect to the intermediate image and producing an output image having a predetermined resolution, and

improving sharpness of the produced output image,

wherein improving the sharpness comprises:

picking up two representative colors in an overlapped block of a predetermined size;

increasing contrast between the picked-up representative colors;

transiting a pixel value existing in an edge region to an approximate representative color of the two contrast-increased representative colors; and

adding result values of the overlapped block by using a Hanning window and calculating a final pixel value in order to remove block discontinuity.

11. The resolution-converting method as claimed in claim 10 , wherein the contrast of the representative colors is adaptively increased through a distance and a dot product between an input pixel and the two representative colors.

12. The resolution-converting method as claimed in claim 10 , wherein the two representative colors are picked up by a K-means algorithm.

13. A resolution-converting apparatus, comprising:

an edge-directed interpolation unit for applying an edge-directed interpolation to an input image and calculating an intermediate image; and

a linear filtering unit for converting a sampling rate with respect to the intermediate image and calculating an output image having a predetermined resolution,

wherein the edge-directed interpolation comprise,

a first interpolation unit to generate interpolated pixels of a part of the input image, and

a second interpolation unit to generate interpolated pixels of other remaining part of the input image.

14. The resolution-converting apparatus as claimed in claim 13 , wherein the edge-directed interpolation unit uses absolute differences of pixels of the input image to estimate an edge direction, adaptively determines an interpolation coefficient based on a pixel value of a position corresponding to the estimated edge direction, uses the edge direction and the interpolation coefficient, calculates the edge-directed interpolation value, adaptively blends an edge-directed interpolation value and an interpolation value by a predetermined linear interpolation method, and calculates a final interpolation value.

15. The resolution-converting apparatus as claimed in claim 14 , wherein the edge direction is estimated with respect to two diagonal directions and four non-diagonal directions.

16. The resolution-converting apparatus as claimed in claim 14 , wherein the predetermined linear interpolation method is a bi-cubic interpolation method.

17. The resolution-converting apparatus as claimed in claim 14 , wherein the edge-directed interpolation unit, if the region has a high spatial frequency, uses an interpolation value by the predetermined linear interpolation method to calculate a final interpolation value.

18. The resolution-converting apparatus as claimed in claim 17 , wherein it is determined whether the region has the high spatial frequency by using mean-deviated local variance in the interpolation region.

19. A resolution-converting apparatus, comprising:

an edge-directed interpolation unit for applying an edge-directed interpolation to an input image and calculating an intermediate image; and

a linear filtering unit for converting a sampling rate with respect to the intermediate image and calculating an output image having a predetermined resolution; and

a sharpness improvement unit for improving sharpness of the calculated output image, wherein the sharpness improvement unit picks up two representative colors in an overlapped block of a predetermined size, increases contrast between the picked-up representative colors, transits a pixel value existing in an edge region to an approximate representative color of the two contrast-increased representative colors, adds result values of the overlapped block by using a Hanning window, and calculates a final pixel value in order to remove block discontinuity.

20. The resolution-converting apparatus as claimed in claim 19 , wherein the contrast of the representative colors is adaptively increased through a distance and a dot product between an input pixel and the two representative colors.

21. The resolution-converting apparatus as claimed in claim 19 , wherein the two representative colors are picked up by the-a K-means algorithm.

22. A resolution-converting method, comprising:

determining whether an interpolation region has a high frequency; and

interpolating using an edge-directed interpolation method and a predetermined linear interpolation method if the interpolation region has not the high frequency.

23. The resolution-converting method as claimed in claim 22 , comprising:

interpolating using only the predetermined linear interpolation method if the interpolation region has the high frequency.

24. The resolution-converting method as claimed in claim 23 , wherein the predetermined linear interpolation method is a bi-cubic interpolation method.