WO2016165097A1 - 一种图像去锯齿的系统 - Google Patents

一种图像去锯齿的系统 Download PDF

Info

Publication number
WO2016165097A1
WO2016165097A1 PCT/CN2015/076720 CN2015076720W WO2016165097A1 WO 2016165097 A1 WO2016165097 A1 WO 2016165097A1 CN 2015076720 W CN2015076720 W CN 2015076720W WO 2016165097 A1 WO2016165097 A1 WO 2016165097A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
edge
filter
image
edge direction
Prior art date
Application number
PCT/CN2015/076720
Other languages
English (en)
French (fr)
Inventor
郭若杉
韩睿
汤仁君
Original Assignee
中国科学院自动化研究所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院自动化研究所 filed Critical 中国科学院自动化研究所
Priority to US15/557,081 priority Critical patent/US10410326B2/en
Priority to PCT/CN2015/076720 priority patent/WO2016165097A1/zh
Publication of WO2016165097A1 publication Critical patent/WO2016165097A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/60—Editing figures and text; Combining figures or text
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20172—Image enhancement details
    • G06T2207/20192—Edge enhancement; Edge preservation

Definitions

  • Processing such as scaling and deinterlacing in image and video processing introduces a sawtooth effect on the edges of the image, ie, the edges that appear straight in the source image appear on the processed image with a sawtooth effect near the edges.
  • the rendered line will be jagged if not handled properly.
  • the sawtooth needs to be removed.
  • the second disadvantage is that when dealing with low angles (ie, angles close to the horizontal direction), for directions below the lowest angle in the quantization direction, the direction cannot be effectively processed. Sawtooth.
  • the third disadvantage is that when dealing with small angles (ie, angles close to the horizontal direction), the pixels participating in the filtering are far apart and are prone to errors.
  • the present invention proposes an image de-saw system which can well handle direction filtering in any direction and has a good effect on low angle saw teeth.
  • the invention provides an image de-saw system, comprising an edge detecting unit, a directional filter and a result fusion unit.
  • the edge detecting unit outputs an edge direction and an edge confidence of the pixel according to the input image
  • the directional filter is based on the edge of the input image and the pixel.
  • the direction output direction is filtered, and the fusion unit outputs an optimized image according to the edge confidence of the pixel, the input image, and the direction filtering result, and the direction filter calculates the edge direction of the pixel according to the edge direction of the pixel to be interpolated.
  • intersection of a line and a horizontal scanning line or a vertical scanning line of the neighborhood and using a one-dimensional interpolation method, inserts a temporary pixel located at the intersection, and then filters each temporary pixel along the edge direction by using a filter. , output direction filtering result.
  • an image de-saw system of the present invention further comprises a small angle direction filter and a direction filtering fusion unit; the small angle direction filter outputs a small angle direction filtering result according to the input image and the edge direction of the pixel.
  • the direction filtering fusion unit weights the direction filter output result and the small angle direction filter output result according to the edge direction of the pixel, and outputs the direction filter weighted combination result; the result fusion unit is weighted according to the edge confidence of the pixel, the input image, and the direction filter. Combining the results to output an optimized image;
  • the directional filter calculates the intersection coordinates of a line along the edge direction of the pixel to be interpolated and the horizontal scanning line of the neighborhood to obtain a temporary pixel located at the intersection, and the small angle direction filter is determined according to the pixel to be interpolated In the edge direction, calculate the intersection coordinates of a line passing through the edge direction of the pixel and the vertical scanning line of the neighborhood, and insert a temporary pixel located at the intersection point by using one-dimensional interpolation; and then use a one-dimensional filter to Each temporary pixel in the edge direction is filtered to output a small angle direction filtering result.
  • the directional filter calculates the intersection coordinates of a line along the edge direction of the pixel to be interpolated and the vertical scanning line of the neighborhood to obtain a temporary pixel located at the intersection, and the small angle direction filter is determined according to the pixel to be interpolated In the edge direction, calculate the intersection coordinates of a line passing along the edge direction of the pixel and the horizontal scanning line of the neighborhood, and insert a temporary pixel located at the intersection point by using one-dimensional interpolation; and then use a one-dimensional filter to Each temporary pixel in the edge direction is filtered to output a small angle direction filtering result.
  • Figure 1 is a block diagram of a conventional anti-aliasing system
  • Figure 2 is a block diagram of an image de-saw system of the present invention
  • Figure 4 is a schematic diagram of the normalized edge confidence calculation curve
  • Figure 5 is a schematic structural view of a directional filter of the present invention.
  • Figure 6 is a schematic diagram of direction filtering of the present invention.
  • Figure 8 is a schematic diagram of the small angle direction filtering of the present invention.
  • Figure 9 is a schematic structural view of a small angle direction filter of the present invention.
  • FIG. 10 is a schematic diagram of a fusion weight calculation curve of the directional filtering of the present invention.
  • the image de-saw system of the embodiment includes an edge detecting unit, a direction filter, a small angle direction filter, a direction filtering fusion unit, and a result fusion unit, and the edge detecting unit outputs the edge direction of the pixel according to the input image f.
  • the direction filter outputs a direction filtering result f dh according to the input image f and the edge direction D of the pixel
  • the small angle direction filter outputs the small angle direction filtering result f according to the input image f and the edge direction D of the pixel.
  • the direction filtering fusion unit weights the direction filter output result f dh and the small angle direction filter output result f dl according to the edge direction D of the pixel, and outputs the direction filter weighted combination result f d ;
  • the result fusion unit is based on the edge of the pixel
  • the degree R edge , the input image f, and the direction filter weighted combination result f d output an optimized image f'.
  • the edge detection unit includes a horizontal gradient calculation module, a vertical gradient calculation module, a local covariance matrix calculation module, an eigenvalue and feature vector calculation module, an edge direction calculation module, an edge confidence calculation module, and a horizontal gradient calculation.
  • the horizontal gradient g x and the vertical gradient g y can be obtained by using a sobel gradient operator or other gradient operators.
  • the formula for calculating the local covariance M c is as shown in equation (1), assuming that the window size used to calculate the local covariance is (2K+1)*(2N+1), then the local covariance of the pixel at (i,j) for
  • Covariance matrix M c eigenvalue ⁇ 1 and 2, ⁇ 1 ⁇ 2, [lambda] is the eigenvector e ⁇ 1 and e ⁇ 2. Then, the edge direction D and the smaller eigenvalue ⁇ 2 correspond to the same eigenvector, that is, as shown in the formula (2)
  • the confidence R of the edge direction is calculated as shown in the equation (3), where Tr represents the trace of the matrix and Det represents the determinant of the matrix.
  • R_T1 and R_T2 are two preset thresholds.
  • the directional filter can achieve the corresponding filtering effect by using the adjacent horizontal scanning line and the adjacent vertical scanning line.
  • This embodiment uses a neighborhood horizontal scanning line for detailed description.
  • the directional filter calculates the intersection coordinates of a line passing along the edge direction of the pixel and the horizontal scanning line of the neighborhood according to the edge direction D of the pixel to be processed, and inserts the intersection point by using a one-dimensional interpolation method.
  • the temporary pixels are filtered by the filter to filter the respective temporary pixels along the edge direction, and the direction filtering result f dh is output.
  • the schematic diagram of the direction filtering is shown in Fig. 6.
  • the black point is the original image pixel
  • P is the pixel to be filtered.
  • a line Lp and a neighborhood level along the edge direction of the pixel are calculated.
  • the coordinates of the intersection of the scan lines (H0, H1, H2, H3, H4 shown in Figure 6), and the horizontal interpolation method insert the temporary pixels (P0, P1, P2, P3, P4) at the intersection, such as As shown by the five-pointed star of Fig. 6, the temporary pixels in the respective edge directions are filtered by a one-dimensional filter along the direction to obtain a final direction filtering result.
  • the coordinates of the intersection P0 between the edge direction and the horizontal scanning line H0 are calculated, and the edge direction and horizontal scanning line intersection calculating unit 4 calculates the coordinates of the intersection P4 between the edge direction and the horizontal scanning line H4.
  • the coordinate calculation process of the intersection point is as follows.
  • the temporary pixel points P0, P1, P3, P4 in the edge direction and the pixel to be filtered The distance between P is far away. As shown in Fig. 7, the pixels far apart may not belong to the edge where the pixel to be filtered is located, so that non-edge pixels also participate in smoothing filtering, and errors may occur.
  • the embodiment adopts a small angle direction filtering method as shown in FIG. 8, and according to the edge direction D of the pixel P to be filtered, a line Lp and a neighborhood along the edge direction of the pixel are calculated at this time.
  • the intersection coordinates of the vertical scanning lines (V0, V1, V2, V3, V4 shown in Fig. 8), and the temporary pixels P0, P1, P2, P3, and P4 located at the intersection point are inserted by the one-dimensional interpolation method (e.g. As shown by the five-pointed star in Fig. 8, each temporary pixel in the edge direction is filtered by a one-dimensional filter, and the small angle direction filtering result f dl is output.
  • the structure diagram of the small-angle direction filter unit is as shown in Fig. 9, and receives the input of the image f and the direction D calculated by the edge detecting unit, and outputs the result of the direction filtering f dl .
  • the present embodiment only exemplifies that the intersection calculation unit of one line passing through the edge direction of the pixel and the adjacent horizontal scanning line is five.
  • the unit 0 calculates the coordinates of the intersection P0 between the edge direction and the vertical scanning line V0
  • the edge direction and vertical scanning line intersection calculating unit 4 calculates the coordinates of the intersection P4 between the edge direction and the vertical scanning line V4.
  • the coordinate calculation process of the intersection point is as follows.
  • the one-dimensional interpolation unit in the small-angle direction filter accepts the image input, and the intersection coordinates calculated by the intersection calculating unit of the edge direction and the vertical scanning line, and inserts the temporary pixel located at the intersection point by the one-dimensional interpolation method.
  • Point Pn One-dimensional interpolation calculates the value of the point to be interpolated using a weighted combination of several original pixels that are adjacent in the vertical direction.
  • the specific one-dimensional interpolation method can select cubic interpolation, linear interpolation, and multi-phase filter interpolation.
  • ⁇ is the angle corresponding to the edge direction D
  • ⁇ _T1 and ⁇ _T2 are the preset two threshold values.
  • the result fusion unit accepts the image input f and the direction filtering result f d , and the edge confidence R edge , and performs weighted fusion on f d and f according to the edge confidence, as shown in equation (12).

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Processing (AREA)

Abstract

本发明公开了一种图像去锯齿系统包括边缘检测单元、方向滤波器、小角度方向滤波器、方向滤波融合单元、结果融合单元，边缘检测单元依据输入图像输出像素的边缘方向和边缘置信度，方向滤波器依据输入图像和像素的边缘方向输出方向滤波结果，小角度方向滤波器依据输入图像和像素的边缘方向输出小角度方向滤波结果；方向滤波融合单元依据各像素的边缘方向将方向滤波器输出结果和小角度方向滤波器输出结果加权组合，输出方向滤波加权组合结果；结果融合单元依据各像素的边缘置信度、输入图像、方向滤波加权组合结果输出优化图像。本发明能很好地处理任意方向的方向滤波，并对低角度锯齿具有良好的效果。

Description

一种图像去锯齿的系统 技术领域
本发明涉及图像视频处理技术领域，尤其涉及对图像、视频、图形进行锯齿去除的领域。
背景技术
图像和视频处理中的缩放和去隔行等处理会给图像的边缘带来锯齿效应，即在源图像中看上去直的边缘在处理后的图像上，在边缘附近出现了锯齿效应。在图形渲染中，处理一条直线时，如果处理不当，渲染出的直线会出现锯齿。当图像出现锯齿时，需要对锯齿进行去除。
一种传统的去锯齿系统如图1所示，先检测图像的边缘，根据边缘方向在方向滤波器组中选取一个滤波器，进行沿边缘方向的滤波，最后利用边缘的置信度对源图像和方向滤波融合单元组合的结果进行组合，输出最终结果。这种去锯齿系统有几个缺点，第一个缺点是方向往往经过量化，量化成有限的几个的方向，每个方向对应一个沿方向的滤波器，这样对于落在量化方向中间的方向，只能选取就近的量化方向进行滤波，或者对邻近的几个量化方向滤波结果进行组合，准确性受到影响。第二个缺点是在处理低角度(即接近水平方向的角度)时，对于低于量化方向中最低角度的方向，无法有效处理该方向出现 的锯齿。第三个缺点是在处理小角度(即接近水平方向的角度)时，参与滤波的像素相隔较远，容易造成错误。
发明内容
为了解决上述技术问题，本发明提出了一种图像去锯齿系统，该系统能很好地处理任意方向的方向滤波，并对低角度锯齿具有良好的效果。
本发明提出的一种图像去锯齿系统，包括边缘检测单元、方向滤波器、结果融合单元，边缘检测单元依据输入图像输出像素的边缘方向和边缘置信度，方向滤波器依据输入图像和像素的边缘方向输出方向滤波结果，结果融合单元依据像素的边缘置信度、输入图像和方向滤波结果输出优化图像，所述的方向滤波器根据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域水平扫描线或邻域垂直扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素，再利用滤波器对沿边缘方向上的各个临时像素进行滤波，输出方向滤波结果。
为了实现更好的去锯齿效果，本发明的一种图像去锯齿系统还包括小角度方向滤波器、方向滤波融合单元；小角度方向滤波器依据输入图像和像素的边缘方向输出小角度方向滤波结果；方向滤波融合单元依据像素的边缘方向将方向滤波器输出结果和小角度方向滤波器输出结果加权组合，输出方向滤波加权组合结果；结果融合单元依据像素的边缘置信度、输入图像、方向滤波加权组合结果输出优化图像；
所述方向滤波器通过计算穿过待插值像素的沿边缘方向的一条线与邻域水平扫描线的交点坐标来得到位于交点的临时像素时，所述的小角度方向滤波器依据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域垂直扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素；再利用一维滤波器对沿边缘方向上的各个临时像素进行滤波，输出小角度方向滤波结果。
所述方向滤波器通过计算穿过待插值像素的沿边缘方向的一条线与邻域垂直扫描线的交点坐标来得到位于交点的临时像素时，所述的小角度方向滤波器依据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域水平扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素；再利用一维滤波器对沿边缘方向上的各个临时像素进行滤波，输出小角度方向滤波结果。
本发明具有如下有益效果
(1)由于不进行方向量化，可以处理任意方向的锯齿；
(2)可以很好地处理小角度的锯齿，可以处理的角度可以任意接近水平方向。
附图说明
图1传统的去锯齿系统框图；
图2本发明图像去锯齿系统框图；
图3本发明边缘检测单元框图；
图4归一化的边缘置信度计算曲线示意图；
图5本发明方向滤波器结构示意图；
图6本发明方向滤波示意图；
图7本发明在小角度进行方向滤波示意图；
图8本发明小角度方向滤波示意图；
图9本发明小角度方向滤波器结构示意图；
图10本发明方向滤波的融合权重计算曲线示意图；
具体实施方式
为使本发明的目的、技术方案和优点更加清楚明白，以下结合具体实施例，并参照附图，对本发明进一步详细说明。
如图2所示，本实施例的图像去锯齿系统包括边缘检测单元、方向滤波器、小角度方向滤波器、方向滤波融合单元、结果融合单元，边缘检测单元依据输入图像f输出像素的边缘方向D和边缘置信度Redge，方向滤波器依据输入图像f和像素的边缘方向D输出方向滤波结果fdh，小角度方向滤波器依据输入图像f和像素的边缘方向D输出小角度方向滤波结果fdl；方向滤波融合单元依据像素的边缘方向D将方向滤波器输出结果fdh和小角度方向滤波器输出结果fdl加权组合，输出方向滤波加权组合结果fd；结果融合单元依据像素的边缘置信度Redge、输入图像f、方向滤波加权组合结果fd输出优化图像f’。
如图3所示，边缘检测单元包括水平梯度计算模块、垂直梯度计算模块、局部协方差矩阵据算模块、特征值和特征向量计算模块、边缘方向计算模块、边缘置信度计算模块，水平梯度计算模块依据输入图像f计算图像像素的水平梯度gx，垂直梯度计算模块依据输入图像f计算图 像像素的垂直梯度gy，局部协方差矩阵据算模块依据像素的水平梯度gx和垂直梯度gy计算像素的局部协方差Mc，特征值和特征向量计算模块依据局部协方差Mc计算像素局部协方差矩阵的特征值λ和特征向量e，边缘方向计算模块依据特征值λ计算像素的边缘方向d，边缘置信度计算模块用于依据特征向量e计算像素的边缘置信度Redge。
其中水平梯度gx和垂直梯度gy可以用sobel梯度算子或其他梯度算子得到。局部协方差Mc的计算公式如式(1)，假设用于计算局部协方差的窗口大小为(2K+1)*(2N+1)，则位于(i,j)的像素其局部协方差为
Figure PCTCN2015076720-appb-000001
协方差矩阵Mc特征值为λ1和λ2，λ1≥λ2，特征向量为eλ1和eλ2。则边缘方向D和较小特征值λ2对应的特征向量相同，即如式(2)所示
D＝eλ2   (2)
边缘方向的置信度R计算如式(3)所示，其中Tr表示矩阵的迹，Det表示矩阵的行列式。
Figure PCTCN2015076720-appb-000002
归一化的边缘置信度Redge计算如图4所示，其公式如图4所示
Figure PCTCN2015076720-appb-000003
其中R_T1和R_T2为预设的两个阈值。
方向滤波器采用邻域水平扫描线和邻域垂直扫描线均可达到相应 的滤波效果，本实施例采用邻域水平扫描线进行详细说明。所述的方向滤波器根据待处理像素的边缘方向D，计算出穿过该像素的沿边缘方向的一条线与邻域水平扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素，再利用滤波器对沿边缘方向上的各个临时像素进行滤波，输出方向滤波结果fdh。
为方便叙述和理解，本实施例仅举穿过该像素的沿边缘方向的一条线与邻域水平扫描线的交点计算单元为5个的例子。
方向滤波示意图如图6所示，其中黑点为原始图像像素，P为待滤波像素，根据待滤波像素P的边缘方向，计算出穿过该像素的沿边缘方向的一条线Lp与邻域水平扫描线(如图6所示的H0，H1,H2,H3,H4)的交点坐标，并利用水平插值的方法，插出位于交点的临时像素(P0,P1,P2,P3,P4)，如图6的五角星所示，再利用沿方向的一维滤波器对各个边缘方向上的临时像素进行滤波，得到最终的方向滤波结果。
如图5所示方向滤波器中的边缘方向与水平扫描线交点计算单元一共有m个(本实施例中m＝5)，其中边缘方向与水平扫描线交点计算单元n(n＝0,1...,m-1)计算出边缘方向与水平扫描线Hn(n＝0,1...,m-1)之间的交点Pn的坐标，即边缘方向与水平扫描线交点计算单元0计算出边缘方向与水平扫描线H0之间的交点P0的坐标，边缘方向与水平扫描线交点计算单元4计算出边缘方向与水平扫描线H4之间的交点P4的坐标。交点坐标计算过程如下，设待滤波的像素P的坐标为(i,j)，设输入方向D＝[v1,v2]T，则穿过该像素的沿边缘方向的一条线Lp与水平线Hn(n＝0,1,2,3,4)的交点Pn(n＝0,1,2,3,4)的坐标(Pn.y,Pn,x)如公式(5) 所示
Figure PCTCN2015076720-appb-000004
如图5所示方向滤波器中的一维插值单元接受图像输入，以及边缘方向与水平扫描线的交点计算单元计算出的交点坐标，用一维插值的方法插出边缘方向上的临时像素点Pn。一维插值利用水平邻近的几个原始像素的加权组合计算出待插值点的值。具体一维插值的方法可以选取三次插值，线性插值，及多相位滤波器插值方法。
如图5所示方向滤波器中的一维滤波单元接受5个位于边缘方向上的临时像素点Pn(n＝0,1,2,3,4)的输入，对5个临时像素点进行一维滤波得到最终的滤波结果fdh，如式(6)所示
Figure PCTCN2015076720-appb-000005
其中cn为预设的进行滤波的系数，可以取做cn＝1/5.即均值滤波。
如果用图6所示的方向滤波方式对图像进行滤波，当像素的边缘方向为小角度时(即接近水平角度)时，边缘方向上的临时像素点P0,P1,P3,P4与待滤波像素P之间距离较远，如图7所示，相距较远的像素可能已经不属于待滤波像素所处的边缘，让非边缘的像素也参与平滑滤波，有可能出现错误。
为解决这个问题，本实施例采用如图8所示的小角度方向滤波方法，根据待滤波像素P的边缘方向D，此时计算出穿过该像素的沿边缘方向 的一条线Lp与邻域垂直扫描线(如图8所示的V0，V1,V2,V3,V4)的交点坐标，并利用一维插值的方法，插出位于交点的临时像素P0、P1、P2、P3、P4(如图8的五角星所示)，再利用一维滤波器对沿边缘方向上的各个临时像素进行滤波，输出小角度方向滤波结果fdl。
小角度方向滤波器单元的结构图如图9所示，接受图像f和边缘检测单元计算的方向D的输入，输出方向滤波的结果fdl。为方便叙述和理解，本实施例仅举本实施例仅举穿过该像素的沿边缘方向的一条线与邻域水平扫描线的交点计算单元为5个的例子。
如图9所示小角度方向滤波器中的边缘方向与垂直扫描线交点计算单元一共有m个(本实施例中m＝5)，其中边缘方向与垂直扫描线交点计算单元n(n＝0,1...,m-1)计算出边缘方向与垂直扫描线Vn(n＝0,1...,m-1)之间的交点Pn的坐标，即边缘方向与垂直扫描线交点计算单元0计算出边缘方向与垂直扫描线V0之间的交点P0的坐标，边缘方向与垂直扫描线交点计算单元4计算出边缘方向与垂直扫描线V4之间的交点P4的坐标。交点坐标计算过程如下，设待滤波的像素P的坐标为(i,j)，设输入方向D＝[v1,v2]T，则穿过该像素的沿边缘方向的一条线Lp与垂直扫描线Vn(n＝0,1,2,3,4)的交点Pn(n＝0,1,2,3,4)的坐标(Pn.y,Pn,x)如公式(7)所示
Figure PCTCN2015076720-appb-000006
如图9所示小角度方向滤波器中的一维插值单元接受图像输入，以及边缘方向与垂直扫描线的交点计算单元计算出的交点坐标，用一维插值的方法插出位于交点的临时像素点Pn。一维插值利用垂直方向上邻近的几个原始像素的加权组合计算出待插值点的值。具体一维插值的方法可以选取三次插值，线性插值，及多相位滤波器插值方法。
如图9所示小角度方向滤波器中的一维滤波单元接受5个位于边缘方向上的临时像素点Pn(n＝0,1,2,3,4)的输入，对5个临时像素点进行一维滤波得到最终的滤波结果fdl，如式(8)所示
Figure PCTCN2015076720-appb-000007
其中cn为预设的进行滤波的系数，可以取做cn＝1/5.即均值滤波。
如图2所示的图像去锯齿系统中的方向滤波融合单元接受方向滤波器的输入fdh和小角度方向滤波器的输入fdl，以及边缘方向D＝[v1,v2]T输入,v1代表方向向量的水平分量，v2代表代表方向向量的垂直分量。根据边缘方向的角度进行加权组合，如式(9)所示，其中w为加权的权重
fd＝(1-w)*fdh+w*fdl     (9)
其中权重计算如图10所示，其计算公式如(10)所示
Figure PCTCN2015076720-appb-000008
其中θ为边缘方向D对应的角度，θ_T1和θ_T2为预设的两个阈值。
边缘方向D对应的角度θ的计算公式如(11)所示
Figure PCTCN2015076720-appb-000009
结果融合单元接受图像输入f以及方向滤波结果fd，以及边缘置信度Redge，根据边缘置信度对fd和f进行加权融合，如式(12)所示
fd＝(1-Redge)*f+Redge*fd      (12)

Claims (11)

  1. 一种图像去锯齿系统，包括边缘检测单元、方向滤波器、结果融合单元，边缘检测单元依据输入图像输出像素的边缘方向和边缘置信度，方向滤波器依据输入图像和像素的边缘方向输出方向滤波结果，结果融合单元依据像素的边缘置信度、输入图像和方向滤波结果输出优化图像，其特征在于，所述的方向滤波器根据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域水平扫描线或邻域垂直扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素，再利用滤波器对沿边缘方向上的各个临时像素进行滤波，输出方向滤波结果。
  2. 如权利要求1所述的一种图像去锯齿系统，其特征在于，还包括小角度方向滤波器、方向滤波融合单元；小角度方向滤波器依据输入图像和像素的边缘方向输出小角度方向滤波结果；方向滤波融合单元依据像素的边缘方向将方向滤波器输出结果和小角度方向滤波器输出结果加权组合，输出方向滤波加权组合结果；结果融合单元依据像素的边缘置信度、输入图像、方向滤波加权组合结果输出优化图像；
    所述方向滤波器通过计算穿过待插值像素的沿边缘方向的一条线与邻域水平扫描线的交点坐标来得到位于交点的临时像素时，所述的小角度方向滤波器依据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域垂直扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素；再利用一维滤波器对沿边缘方向上的各个临 时像素进行滤波，输出小角度方向滤波结果；
    所述方向滤波器通过计算穿过待插值像素的沿边缘方向的一条线与邻域垂直扫描线的交点坐标来得到位于交点的临时像素时，所述的小角度方向滤波器依据待插值像素的边缘方向，计算出穿过该像素的沿边缘方向的一条线与邻域水平扫描线的交点坐标，并利用一维插值的方法，插出位于交点的临时像素；再利用一维滤波器对沿边缘方向上的各个临时像素进行滤波，输出小角度方向滤波结果。
  3. 如权利要求2所述的一种图像去锯齿系统，其特征在于，边缘检测单元包括水平梯度计算模块、垂直梯度计算模块、局部协方差矩阵据算模块、特征值和特征向量计算模块、边缘方向计算模块、边缘置信度计算模块，水平梯度计算模块依据输入图像计算图像像素的水平梯度，垂直梯度计算模块依据输入图像计算图像像素的垂直梯度，局部协方差矩阵据算模块依据像素的水平梯度和垂直梯度计算像素的局部协方差，特征值和特征向量计算模块依据局部协方差计算像素局部协方差矩阵的特征值和特征向量，边缘方向计算模块依据特征值计算像素的边缘方向，边缘置信度计算模块用于依据特征向量计算像素的边缘置信度。
  4. 如权利要求3所述的一种图像去锯齿系统，其特征在于，局部协方差的计算公式为
    Figure PCTCN2015076720-appb-100001
    其中局部协方差的窗口大小为(2K+1)*(2N+1)，Mc(i，j)为坐标为(i,j)的像素的局部协方差，gx为像素的水平梯度，gy为像素的垂直 梯度；
    边缘方向的计算公式为D＝eλ2，其中D为边缘方向，eλ2为协方差矩阵Mc的较小特征值对应的特征向量；
    边缘置信度的计算公式为
    Figure PCTCN2015076720-appb-100002
    其中Redge为归一化的边缘方向的置信度，R_T1和R_T2为预设的两个阈值，R为边缘方向置信度，其中
    Figure PCTCN2015076720-appb-100003
    λ1和λ2分别为协方差矩阵Mc(i,j)的两个特征值为，且λ1≥λ2。
  5. 如权利要求4所述的一种图像去锯齿系统，其特征在于，所述的一维插值的方法可以采用三次插值方法、线性插值方法或多相位滤波器插值方法。
  6. 如权利要求5所述的一种图像去锯齿系统，其特征在于，待处理像素的沿边缘方向的一条线与邻域水平扫描线交点坐标的计算公式为：
    Figure PCTCN2015076720-appb-100004
    其中(Pn.y,Pn,x)为坐标为(i,j)的待处理像素P的边缘方向与水平扫描线交点，m为大于等于2的自然数，v1代表边缘方向向量的水平分量，v2代表代表边缘方向向量的垂直分量。
  7. 如权利要求5所述的一种图像去锯齿系统，其特征在于，待处理像素的沿边缘方向的一条线与邻域垂直扫描线交点坐标的计算公式为：
    Figure PCTCN2015076720-appb-100005
    其中(Pn.y,Pn,x)为坐标为(i,j)的待处理像素P的边缘方向与垂直扫描线交点，m为大于等于2的自然数，v1代表边缘方向向量的水平分量，v2代表代表边缘方向向量的垂直分量。
  8. 如权利要求6或7所述的一种图像去锯齿系统，其特征在于，对各个临时像素进行滤波的滤波器为一维滤波器。
  9. 如权利要求8所述的一种图像去锯齿系统，其特征在于，方向滤波加权组合结果为fd＝(1-w)*fdh+w*fdl，其中fd为方向滤波加权组合结果，w为加权组合权重，fdh方向滤波结果，fdl为小角度方向滤波结果。
  10. 如权利要求9所述的一种图像去锯齿系统，其特征在于，方向滤波融合单元中的加权组合权重w的计算公式为
    Figure PCTCN2015076720-appb-100006
    其中θ为边缘方向D对应的角度，θ_T1和θ_T2为预设的两个阈值。
  11. 如权利要求10所述的一种图像去锯齿系统，其特征在于，优化图像的加权融合公式为
    Figure PCTCN2015076720-appb-100007
    其中f′为优化图像，Redge为归一化的边缘方向的置信度，f为输入图像，fd为方向滤波加权组合结果。
PCT/CN2015/076720 2015-04-16 2015-04-16 一种图像去锯齿的系统 WO2016165097A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US15/557,081 US10410326B2 (en) 2015-04-16 2015-04-16 Image anti-aliasing system
PCT/CN2015/076720 WO2016165097A1 (zh) 2015-04-16 2015-04-16 一种图像去锯齿的系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2015/076720 WO2016165097A1 (zh) 2015-04-16 2015-04-16 一种图像去锯齿的系统

Publications (1)

Publication Number Publication Date
WO2016165097A1 true WO2016165097A1 (zh) 2016-10-20

Family

ID=57125623

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/076720 WO2016165097A1 (zh) 2015-04-16 2015-04-16 一种图像去锯齿的系统

Country Status (2)

Country Link
US (1) US10410326B2 (zh)
WO (1) WO2016165097A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107492073A (zh) * 2017-07-14 2017-12-19 天津大学 基于权重引导滤波的数字图像修复算法
CN112651304A (zh) * 2020-12-11 2021-04-13 西安电子科技大学 基于特征融合的可撤销掌纹模板生成方法、装置、设备和存储介质

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109934789B (zh) * 2019-03-26 2021-01-01 湖南国科微电子股份有限公司 图像去噪方法、装置及电子设备
CN110827220B (zh) * 2019-11-01 2023-05-30 杭州当虹科技股份有限公司 一种基于图像边缘分析的抗锯齿方法
TWI774224B (zh) * 2021-02-03 2022-08-11 緯創資通股份有限公司 特徵強化與資料擴增方法及其動作偵測裝置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020126900A1 (en) * 2001-01-05 2002-09-12 Sang Yeon Kim Image interpolation method and apparatus thereof
US20030206667A1 (en) * 2003-05-30 2003-11-06 Samsung Electronics Co., Ltd. Edge direction based image interpolation method
CN101790069A (zh) * 2010-03-09 2010-07-28 周艇 一种基于图像边缘方向的尺度变换方法
CN103093418A (zh) * 2013-02-21 2013-05-08 深圳市晶日盛科技有限公司 一种改进的数字图像缩放方法
CN103500435A (zh) * 2013-09-11 2014-01-08 西安交通大学 一种边缘导向的自适应图像插值方法及其vlsi实现装置
CN104794692A (zh) * 2015-04-16 2015-07-22 中国科学院自动化研究所 一种图像去锯齿的系统

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4710635B2 (ja) * 2006-02-07 2011-06-29 ソニー株式会社 画像処理装置および方法、記録媒体、並びに、プログラム
US8559746B2 (en) * 2008-09-04 2013-10-15 Silicon Image, Inc. System, method, and apparatus for smoothing of edges in images to remove irregularities
US8180180B2 (en) * 2008-12-29 2012-05-15 Arcsoft Hangzhou Co., Ltd. Method for magnifying images and videos
US8391628B2 (en) * 2010-12-16 2013-03-05 Stmicroelectronics Asia Pacific Pte Ltd. Directional anti-aliasing filter
US9336567B2 (en) * 2013-12-16 2016-05-10 Telefonaktiebolaget L M Ericsson (Publ) Content-aware weighted image manipulations
US20170302910A1 (en) * 2016-04-19 2017-10-19 Motorola Mobility Llc Method and apparatus for merging depth maps in a depth camera system

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020126900A1 (en) * 2001-01-05 2002-09-12 Sang Yeon Kim Image interpolation method and apparatus thereof
US20030206667A1 (en) * 2003-05-30 2003-11-06 Samsung Electronics Co., Ltd. Edge direction based image interpolation method
CN101790069A (zh) * 2010-03-09 2010-07-28 周艇 一种基于图像边缘方向的尺度变换方法
CN103093418A (zh) * 2013-02-21 2013-05-08 深圳市晶日盛科技有限公司 一种改进的数字图像缩放方法
CN103500435A (zh) * 2013-09-11 2014-01-08 西安交通大学 一种边缘导向的自适应图像插值方法及其vlsi实现装置
CN104794692A (zh) * 2015-04-16 2015-07-22 中国科学院自动化研究所 一种图像去锯齿的系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107492073A (zh) * 2017-07-14 2017-12-19 天津大学 基于权重引导滤波的数字图像修复算法
CN112651304A (zh) * 2020-12-11 2021-04-13 西安电子科技大学 基于特征融合的可撤销掌纹模板生成方法、装置、设备和存储介质

Also Published As

Publication number Publication date
US20180061013A1 (en) 2018-03-01
US10410326B2 (en) 2019-09-10

Similar Documents

Publication Publication Date Title
WO2016165097A1 (zh) 一种图像去锯齿的系统
CN104794692B (zh) 一种图像去锯齿的系统
US9654765B2 (en) System for executing 3D propagation for depth image-based rendering
CN105100640B (zh) 一种局部配准并行视频拼接方法及系统
JP4727720B2 (ja) 画像処理方法および画像処理装置
CN104700360B (zh) 基于边缘自适应的图像缩放方法及系统
WO2012165491A1 (ja) ステレオカメラ装置、および、コンピュータ読み取り可能な記録媒体
US8396330B2 (en) Image upscaling based upon directional interpolation
US7751641B2 (en) Method and system for digital image enhancement
US8494308B2 (en) Image upscaling based upon directional interpolation
US8045053B2 (en) Video image deinterlacing apparatus and methods of performing video image deinterlacing
CN108445009B (zh) 一种太阳能电池板裂纹检测方法
CN116165216B (zh) 一种液晶显示屏微划伤瑕疵3d检测方法、系统及计算设备
WO2014127665A1 (zh) 一种改进的数字图像缩放方法
CN105678719A (zh) 一种全景拼接接缝处平滑方法及装置
CN110349090B (zh) 一种基于牛顿二阶插值的图像缩放方法
JP2009044594A5 (zh)
US20160335517A1 (en) Line segment and arc detection apparatus
Priyadarshini et al. A new edge detection method based on additions and divisions
TW201536029A (zh) 影像縮減取樣裝置與方法
JP2009094862A (ja) 映像信号処理装置
TWI429274B (zh) 具有縮放與清晰裝置之影像處理系統及其方法
Woetzel et al. Multi-Camera Real-Time Depth Estimation with Discontinuity Handling on PC Graphics Hardware.
CN103258318B (zh) 一种图像降噪处理方法及系统
JP5995704B2 (ja) 画像処理装置及び画像処理方法並びにプログラム及び記録媒体

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15888802

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 15557081

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15888802

Country of ref document: EP

Kind code of ref document: A1