Method for anti-aliasing of image with super-resolution

Patent metadata

Field	Value
Publication number	US-2016247262-A1
Application number	US-201414380751-A
Country	US
Kind code	A1
Filing date	May 15, 2014
Priority date	Apr 28, 2014
Publication date	Aug 25, 2016
Grant date	—

Abstract

A method for image edge anti-aliasing with super-resolution is provided and used to achieve enlargement conversion from low resolution to high resolution, by applying the steps of: detecting a received image data and saving as an original edge pixel frame; enlarging the original edge pixel frame in double size along the horizontal and vertical directions, respectively; retaining the original edge pixel information; replacing pixels to be interpolated with a zero grayscale; and compensating the pixels to be interpolated which are temporarily replaced by the zero grayscale along edge directions of the original edge pixel, such that the jagged phenomenon of output picture is output picture is significantly decreased, such that the jagged phenomena of an output picture is significantly decreased, with a detailed image information is well-maintained. The method is simple, with fewer calculations and faster operating speed, the cost can be effectively reduced.

First claim

What is claimed is:

1 . A method of anti-aliasing of an image with super-resolution comprising the steps of:
(1) adopting a Canny edge detection to detect edges of an image and saving the edges as an original edge pixel frame, which comprises the steps of: smoothing the image edges by using a Gaussian filter; calculating a gradient magnitude and a gradient direction of the edges by applying a first-order partial derivative of the finite difference; adopting an non-maxima suppression to the gradient magnitude of the edges; and applying a dual-threshold algorithm to detect and connect the edges; (2) enlarging the original edge pixel frame to form a 2×2 enlarged pixel frame, retaining all information of edge pixels of the original edge pixel frame, additional pixels of the enlarged pixel frame caused by the enlargement are pixels to be interpolated; (3) replacing the pixels to be interpolated with a zero grayscale; and (4) compensating the pixels to be interpolated which are temporarily replaced by the zero grayscale along edge directions of the retained original edge pixels in the enlarged pixel frame, which comprises the steps of: dividing the edge pixels detected on the basis of the step (1) into a plurality of 4×4 pixels, and determining whether there is a special edge within the 4×4 pixel, if so, applying a corresponding 4×4 interpolation rule to calculate, and if not, determining whether there is a 2×2 universal edge within the 4×4 pixel, if so, applying a 2×2 common interpolation rule to compensate the zero grayscale pixels, if the above two edges do not exist, then applying a double-cubic interpolation algorithm to compensate the interpolation of the zero grayscale pixels; wherein the 4×4 interpolation rule is configured to determine an edge direction of the 4×4 pixel, and interpolating along the edge direction based on a center point of the 4×4 pixel, the interpolations of a top point and a left point of the center point are simultaneously compensated; the 2×2 common interpolation rule is for determining the edge direction of only four pixel points in the center of the 4×4 pixel, and interpolating along the edge direction based on the center point of the 4×4 pixel, the interpolations of the top point and the left point of the center point are simultaneously compensated; the compensation for interpolating of the center point of the 4×4 pixel, the top point, and the left point of the center point are determined by the four pixel points in the center of the 4×4 pixel.

2 . A method of anti-aliasing of an image with super-resolution comprising the steps of:
(1) detecting edges of an image and saving the edges as an original edge pixel frame; (2) enlarging the original edge pixel frame to form a 2×2 enlarged pixel frame, retaining all the information of edge pixels of the original edge pixel frame, additional pixels of the enlarged pixel frame caused by the enlargement are pixels to be interpolated; (3) replacing the pixels to be interpolated with a zero grayscale; (4) compensating the pixels to be interpolated which are temporarily replaced by the zero grayscale along edge directions of the retained original edge pixels in the enlarged pixel frame.

3 . The method as claimed in claim 2 , wherein the step (1) comprises:
(1-1) smoothing the image edges by using a Gaussian filter; (1-2) calculating a gradient magnitude and a gradient direction of the edges by applying a first-order partial derivative of the finite difference; (1-3) adopting an non-maxima suppression to the gradient amplitude of the edges; and (1-4) applying a dual-threshold algorithm to detect and connect the edges.

4 . The method as claimed in claim 2 , wherein the step (4) comprises: dividing the edge pixels detected on the basis of the step (1) into a plurality of 4×4 pixels, and determining whether there is a special edge within the 4×4 pixel, if so, applying a corresponding 4×4 interpolation rule to calculate, and if not, determining whether there is a 2×2 universal edge within the 4×4 pixel, if so, applying a 2×2 common interpolation rule to compensate the zero grayscale pixels, if the above two edges do not exist, then applying a double-cubic interpolation algorithm to compensate the interpolation of the zero grayscale pixels.

5 . The method as claimed in claim 4 , wherein the 4×4 interpolation rule is configured to determine an edge direction of the 4×4 pixel, the interpolations of top point and left point of the center point are simultaneously compensated.

6 . The method as claimed in claim 4 , wherein the 2×2 common interpolation rule is for determining the edge direction of only four pixel points in the center of the 4×4 pixel, and interpolating along the edge direction based on the center point of the 4×4 pixel, the interpolations of the top point and the left point of the center point are simultaneously compensated.

7 . The method as claimed in claim 5 , wherein the compensation for interpolating of the center point of the 4×4 pixel, the top point and the left point of the center point are determined by the four pixel points in the center of the 4×4 pixel.

8 . The method as claimed in claim 6 , wherein the compensation for interpolating of the center point of the 4×4 pixel, the top point and the left point of the center point are determined by the four pixel points in the center of the 4×4 pixel.

9 . The method as claimed in claim 7 , wherein a pixel coordinate of a pixel located in lower left corner of the 4×4 pixel is defined as (0, 0), and a pixel coordinate of a pixel located in upper right corner is defined as (3a, 3b), when pixel coordinates (0, b), (a, b), (2a, b), (3a, b) are detected as edge pixel points, a pixel value of the center point of the 4×4 pixel is derived by bicubic interpolation of four pixel coordinates (0, a), (a, b), (2a, b), (3a, b), a pixel value of the left point of the center point is equal to that of the pixel coordinate (a, b), a pixel value of the top point of the center point is equal to an average pixel value of pixel coordinates (a, 2b), (2a, 2b); when pixel coordinates (0, 2b), (a, 2b), (2a, 2b), (3a, 2b) are detected as edge pixel points, a pixel value of the top point of the center point is derived by bicubic interpolation of the four pixel coordinates (0, 2b), (a, 2b), (2a, 2b), (3a, 2b), a pixel value of the left point of the center point is equal to that of the pixel coordinate (a, b), a pixel value of the center point of the 4×4 pixel is equal to an average pixel value of the pixel coordinates (a, b), (2a, b), where a represents a unit length in a direction of the X-axis, b represents a unit length in a direction of the Y-axis.

10 . The method as claimed in claim 7 , wherein a pixel coordinate of a pixel located in lower left corner of the 4×4 pixel is defined as (0, 0), and a pixel coordinate of a pixel located in upper right corner is defined as (3a, 3b), when pixel coordinates (0, b), (a, b), (2a, 2b), (3a, 2b), and (0, 2b), (a, 2b), (2a, b), (3a, b) are respectively detected as edge pixel points, a pixel value of the center point of the 4×4 pixel is equal to an average pixel value of the pixel coordinates (a, b), (2a, b), a pixel value of the left point of the center point is equal to that of the pixel coordinate (a, 2b), a pixel value of the top point of the center point is equal to that of the pixel coordinates (2a, 2b) and (a, 2b) respectively, where a represents a unit length in a direction of the X-axis, b represents a unit length in a direction of the Y-axis.

11 . The method as claimed in claim 7 , wherein a pixel coordinate of a pixel located in lower left corner of the 4×4 pixel is defined as (0, 0), and a pixel coordinate of a pixel located in up

Assignees

• Shenzhen China Star Optoelect

Inventors

• Li Hao
• Hu Houliang
• Chu Li-Wei
• He Jhenwei

Classifications

• G06T3/4007

  based on interpolation, e.g. bilinear interpolation (image demosaicing G06T3/4015; edge-driven or edge-based scaling G06T3/403) · CPC title

• G06T2207/20192

  Edge enhancement; Edge preservation · CPC title

• G06T3/4053

  based on super-resolution, i.e. the output image resolution being higher than the sensor resolution · CPC title

• G06T5/002 Primary

  Physics · mapped topic

• G06T3/403

  Edge-driven scaling; Edge-based scaling · CPC title