Method for motion estimation using a low-bit edge image

• US 20040081240A1
• Filed: 10/29/2002
• Published: 04/29/2004
• Est. Priority Date: 10/29/2002
• Status: Active Grant

• Current Assignee: Winbond Electronics Corporation
• Original Assignee: Winbond Electronics Corporation
• Inventors: Qiu, Xiaohai, Chen, Hongyi
• Granted Patent: US 7,151,798 B2
• Time in Patent Office: Days
• US Class Current: 375/240.16
• CPC Class Codes: H04N 19/51 Motion estimation or motion compensation
• CPC Class Codes: H04N 5/145 Movement estimation for video coding H04N19/51

Abstract

The invention provides a method for a motion estimation algorithm. The motion estimation algorithm using low bit resolution integrated edge image instead of luminance image to obtain difference block with small AC coefficients, edge image created by filters is employed to improve encoding quality, on the other hand, operation cost is reduced to low bit resolution. The invention also provides a method for a motion estimation algorithm with a new algorithm using low-bit resolution oriented edge image. Using low-bit resolution oriented edge in motion estimation can result in flatter image blocks which is in demanded by texture-compress unit (such as DCT), as a result, the encoding efficiency is improved, and the operation load is reduced by low-bit resolution.

Claims

1. A method for motion estimation, comprising:

generating an oriented edge image by an edge-detect-filtering process in according with contour and texture of an original luminance image;

generating an subsampled edge image by performing an low-bit resolution oriented edge image process upon the edge image;

performing a wide-range search upon the subsampled edge image to generate a motion vector;

performing a small-range full search is performed upon the generated motion vector to obtain a final motion vector by using decoded luminance pixels.

2. The method for motion estimation of claim 1, wherein the edge-detect-filtering process including:

performing a x-direction filtering process and a y-direction filtering process to generate a x-direction oriented edge pixels and a y-direction oriented edge pixels; and

generating the oriented edge image by adding the x-direction edge pixels and the y-direction edge pixels.

3. The method for motion estimation of claim 1, wherein

the x-direction edge pixels and the y-direction edge pixels are generated in accordance with following equations, respectively:

$\begin{array}{c}x=\underset{k=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\underset{l=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br><br><br>\mathrm{Cgx}<br><br>\left(x+k,y+1\right)-\mathrm{Rgx}<br><br>\left(x+i+k,y+j+l\right)<br><br>\\ y=\underset{k=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\underset{l=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br><br><br>\mathrm{Cgy}<br><br>\left(x+k,y+1\right)-\mathrm{Rgy}<br><br>\left(x+i+k,y+j+l\right)<br><br>\end{array}$

where Cgx/Cgy and Rgx/Rgy are Gx/Gy edge intensity of a current block and a reference block respectively;

SADgxy(i,j)=SADgx+SADgy=x+y

SADgxy is the sum of absolute difference of the original luminance image and is used for a search criterion.

4. The method for motion estimation of claim 3, wherein the low-bit resolution oriented edge image process including performing a non-uniform quantization process upon the edge image, the non-uniform quantization process including:

providing a plurality of quantization levels in accordance with a predetermined bit-length; and

quantizing the x-direction edge pixels and the y-direction edge pixels in accordance with the quantization levels to generate the subsampled edge image.

5. The method for motion estimation of claim 3, wherein the bit-length is 8-bit.

6. The method for motion estimation of claim 3, wherein the bit-length is 6-bit.

7. The method for motion estimation of claim 3, wherein the bit-length is 4-bit.

8. The method for motion estimation of claim 3, wherein the bit-length is depended upon encoding quality, wherein the more the bit-length, the better the encoding quality.

9. A method for motion estimation, comprising:

generating an oriented edge image by an edge-detect-filtering process in according with contour and texture of an original luminance image, the edge-detect-filtering process including;

performing a x-direction filtering process and a y-direction filtering process to generate a x-direction oriented edge pixels and a y-direction oriented edge pixels; and

generating the oriented edge image by adding the x-direction edge pixels and the y-direction edge pixels;

generating an subsampled edge image by performing an low-bit resolution oriented edge image process upon the edge image, the low-bit resolution oriented edge image process including performing a non-uniform quantization process upon the edge image, the non-uniform quantization process including;

providing a plurality of quantization levels in accordance with a predetermined bit-length, wherein the bit-length is depended upon encoding quality, the more the bit-length, the better the encoding quality; and

quantizing the x-direction edge pixels and the y-direction edge pixels in accordance with the quantization levels to generate the subsampled edge image; and

using the subsampled edge image to generate a motion vector for the original luminance image.

10. The method for motion estimation of claim 9, wherein

the x-direction edge pixels and the y-direction edge pixels are generated in accordance with following equations, respectively:

$\begin{array}{c}x=\underset{k=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\underset{l=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br><br><br>\mathrm{Cgx}<br><br>\left(x+k,y+1\right)-\mathrm{Rgx}<br><br>\left(x+i+k,y+j+l\right)<br><br>\\ y=\underset{k=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>\underset{l=0}{\overset{N-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br><br><br>\mathrm{Cgy}<br><br>\left(x+k,y+1\right)-\mathrm{Rgy}<br><br>\left(x+i+k,y+j+l\right)<br><br>\end{array}$

where Cgx/Cgy and Rgx/Rgy are Gx/Gy edge intensity of a current block and a reference block respectively;

SADgxy(i,j)=SADgx+SADgy=x+y

SADgxy is the sum of absolute difference of the original luminance image and is used for a search criterion.