CN104751177B - A kind of method for being used in reference numbers image give fuzzy digit straightway - Google Patents
A kind of method for being used in reference numbers image give fuzzy digit straightway Download PDFInfo
- Publication number
- CN104751177B CN104751177B CN201510137710.3A CN201510137710A CN104751177B CN 104751177 B CN104751177 B CN 104751177B CN 201510137710 A CN201510137710 A CN 201510137710A CN 104751177 B CN104751177 B CN 104751177B
- Authority
- CN
- China
- Prior art keywords
- straight line
- bdss
- digital
- coordinates
- line segment
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 9
- 230000009466 transformation Effects 0.000 claims description 18
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000012417 linear regression Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 13
- 238000004458 analytical method Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 230000000007 visual effect Effects 0.000 description 5
- 238000002715 modification method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 239000007787 solid Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 238000007792 addition Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000012015 optical character recognition Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000003466 welding Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
本发明涉及一种用于标识数字图像中给定模糊数字直线段的方法，当模糊数字直线段BDSS识别算法完成对数字图像中模糊数字直线段BDSS的识别，模糊数字直线段BDSS以集合的形式保存为识别结果；根据识别结果给定的一个模糊数字直线段BDSS，计算其几何直线特性，使用斜率符合该几何直线特性的矩形边框标识给定像素集合。本发明特别适用于可视化标识数字直线段，模糊数字直线段等数字图像中具有直线几何特征的像素集合，除在数字图像中添加矩形边框外，不修改数字图像中其他原始信息，背景更容易观察；实施简单，耗时与分辨率基本呈线性稳定增加，性能高效。
The invention relates to a method for identifying a given fuzzy digital straight line segment in a digital image. When the fuzzy digital straight line segment BDSS recognition algorithm completes the recognition of the fuzzy digital straight line segment BDSS in the digital image, the fuzzy digital straight line segment BDSS is in the form of a set Save it as the recognition result; calculate its geometric straight line characteristics according to a fuzzy digital straight segment BDSS given by the recognition result, and use a rectangular border whose slope conforms to the geometric straight line characteristics to identify a given set of pixels. The invention is especially suitable for visually identifying digital straight line segments, fuzzy digital straight line segments, and other pixel sets with straight line geometric features in digital images. Except for adding a rectangular frame to the digital image, it does not modify other original information in the digital image, and the background is easier to observe. ; The implementation is simple, the time consumption and resolution basically increase linearly and stably, and the performance is efficient.
Description
技术领域technical field
本发明涉及一种用于标识数字图像中给定模糊数字直线段的方法，即由分散或连续像素组成的集合，进行可视化标识的一种计算方法。The invention relates to a method for identifying a given fuzzy digital straight line segment in a digital image, that is, a calculation method for visually identifying a set composed of scattered or continuous pixels.
背景技术Background technique
数字图像处理中，识别图像中具有特定属性的对象具有很高的学术和应用价值，例如人脸识别，光学字符识别等。数字直线段及衍生的，具有直线几何特征的图形识别属于图像识别领域中的重要分支，其应用包括卫星图片中的道路识别(S.Aliana,V.A.Tolpekina,W.Bijkera,L.Kumarb,“Identifying curvature of overpass mountainroads in Iran from high spatial resolution remote sensing data”,Int.J.ofAppl.Earth Observation and Geoinformation,vol.26,pp.21–25,2014.)，焊接中的焊缝识别(L.Jia,N.Sun,“A line segment detection algorithm based on statisticalanalyses of quantified directions in digital image”,Comput.Modelling&NewTechnologies,vol.18,no.6,pp.79-88,2014.)等。对于数字直线段及衍生对象的自动识别计算方法在国内外学术界不断推陈出新，虽然识别精度和效率逐步上升，但一直采用较粗糙的方法标识结果，如附图说明中图1所示。图1展示了三种常见标识方法，从左至右依次为端点标识法(P.Bhowmick and B.B.Bhattacharya,“Fast polygonal approximation ofdigital curves using relaxed straightness properties”,IEEE Trans.PatternAnal.Mach.Intell.,vol.29,no.9,pp.1590-1602,Sept.2007.)，彩色线段标识法(L.Jia,N.Sun,“A line segment detection algorithm based on statistical analyses ofquantified directions in digital image”,Comput.Modelling&New Technologies,vol.18,no.6,pp.79-88,2014.)和斜率覆盖标识法(L.Buzer,“A simple algorithm fordigital line recognition in the general case”,Pattern Recognition,vol.40,no.6,pp:1675-1684,Jun.2007.)。这三种标识法的共同缺点是使用单色完全覆盖标识对象，过多修改了图像原始数据，导致背景信息丢失，无法清楚标识叠加区域，对人眼观察被识别对象造成了障碍，特别是当识别对象只包含覆盖区域的部分像素时，这种对象包含部分像素的情况是现实图像中的常见情况。学术界常以类似图1中(a)和(c)的相应原始图像开展测试，在这种理想状态下，单色覆盖引起的问题不显著，但涉及到实际应用时，往往需要根据实际图像测试结果反复开展实验，类似图1所示的三种标识法在实际实验中会导致标识区域背景信息的完全丢失且难以区分叠加区域，使得测试结果难以用人眼进行评估。因此，针对图形识别，特别是数字直线段及衍生对象的识别，缺乏一种既实施简单，又能相对完好保存图像背景的标识计算方法。In digital image processing, identifying objects with specific attributes in an image has high academic and application value, such as face recognition, optical character recognition, etc. Digital straight line segment and its derivatives, graphic recognition with straight line geometric features belong to an important branch in the field of image recognition, and its applications include road recognition in satellite images (S.Aliana, V.A.Tolpekina, W.Bijkera, L.Kumarb, "Identifying Curvature of overpass mountainroads in Iran from high spatial resolution remote sensing data”, Int.J.ofAppl.Earth Observation and Geoinformation,vol.26,pp.21–25,2014.), weld recognition in welding (L.Jia , N. Sun, "A line segment detection algorithm based on statistical analyzes of quantified directions in digital image", Comput. Modeling & New Technologies, vol.18, no.6, pp.79-88, 2014.) etc. The calculation methods for automatic recognition of digital straight line segments and derived objects are constantly being innovated in academic circles at home and abroad. Although the recognition accuracy and efficiency are gradually increasing, a rough method has been used to mark the results, as shown in Figure 1 in the accompanying drawings. Figure 1 shows three common identification methods, from left to right the endpoint identification method (P.Bhowmick and B.B.Bhattacharya, "Fast polygonal approximation of digital curves using relaxed straightness properties", IEEE Trans.PatternAnal.Mach.Intell., vol .29, no.9, pp.1590-1602, Sept.2007.), color line segment identification method (L. Jia, N. Sun, "A line segment detection algorithm based on statistical analyzes of quantified directions in digital image", Comput .Modelling&New Technologies, vol.18, no.6, pp.79-88, 2014.) and slope coverage identification method (L.Buzer, "A simple algorithm for digital line recognition in the general case", Pattern Recognition, vol.40 , no.6, pp:1675-1684, Jun.2007.). The common disadvantages of these three marking methods are that they use monochrome to completely cover the marked object, modifying the original image data too much, resulting in the loss of background information, unable to clearly mark the superimposed area, and causing obstacles to the human eye to observe the recognized object, especially when Recognize when an object contains only some pixels of the covered area, which is a common situation in real-world images. The academic community often conducts tests with the corresponding original images similar to (a) and (c) in Figure 1. In this ideal state, the problems caused by monochrome coverage are not obvious, but when it comes to practical applications, it is often necessary to use the actual image Test results Repeated experiments show that the three marking methods shown in Figure 1 will completely lose the background information of the marked area and make it difficult to distinguish the superimposed area in the actual experiment, making it difficult to evaluate the test results with the human eye. Therefore, for graphic recognition, especially the recognition of digital straight line segments and derived objects, there is a lack of a logo calculation method that is simple to implement and can relatively well preserve the image background.
发明内容Contents of the invention
本发明要解决的技术问题是：为了克服现有数字直线段及衍生对象的自动识别后的标识法使用单色完全覆盖标识对象，过多修改了图像原始数据，导致背景信息丢失，无法清楚标识叠加区域，对人眼观察被识别对象造成了障碍的不足，本发明提供一种用于标识数字图像中给定模糊数字直线段的方法，特别适用于可视化标识数字直线段，模糊数字直线段等数字图像中具有直线几何特征的像素集合，除在数字图像中添加矩形边框外，不修改数字图像中其他原始信息。The technical problem to be solved by the present invention is: in order to overcome the existing marking method of digital straight line segments and derived objects after automatic recognition, using a single color to completely cover the marking object, the original data of the image is modified too much, resulting in the loss of background information and the inability to clearly mark The superimposed area causes obstacles to the human eye to observe the identified object. The present invention provides a method for marking a given fuzzy digital straight line segment in a digital image, which is especially suitable for visually identifying digital straight line segments, fuzzy digital straight line segments, etc. A collection of pixels with linear geometric features in a digital image, except for adding a rectangular frame to the digital image, without modifying other original information in the digital image.
为使陈述清楚明了，现集中定义本发明所涉及的部分符号和概念。In order to make the statement clear, some symbols and concepts involved in the present invention are now focused on defining.
Z+表示正整数集合。Z + represents a set of positive integers.
Z表示包括零的整数集合。Z represents the set of integers including zero.
R+表示包括零的正实数集合。R + represents the set of positive real numbers including zero.
R表示包括零的实数集合。R represents the set of real numbers including zero.
max{元素|元素满足的条件}表示满足条件的“最大”元素。max {element | condition satisfied by element} indicates the "maximum" element that satisfies the condition.
min{元素|元素满足的条件}表示满足条件的“最小”元素。min{Element|Condition Satisfied by Element} means the "minimum" element that satisfies the condition.
对于r∈R+， For r∈R + ,
表示数字图像的宽。 Indicates the width of the digital image.
表示数字图像的高。 Indicates the height of the digital image.
存在r1,r2∈R+，满足条件 There exists r 1 , r 2 ∈ R + , satisfying the condition
索引值： index value:
图像数据：以索引值标识并保存的数字化信息。Image data: digitized information identified and saved by index value.
图像空间(记作)：保存数字图像的图像数据。每一个记录单元唯一对应一个像素。图像数据中，对应图像信息左上角的像素索引值为按照对应图像信息的，先从左向右，后从上至下的顺序排列像素，顺序中的第k∈Z+个像素的索引值为 Image space (denoted as ): Saves the image data of a digital image. Each recording unit uniquely corresponds to a pixel. In the image data, the pixel index corresponding to the upper left corner of the image information is According to the corresponding image information, the pixels are arranged in order from left to right, and then from top to bottom, and the index value of the k∈Z + th pixel in the order is
pi表示图像空间中索引值为i的像素。p i represents the pixel with index value i in the image space.
p(x，y)表示图像空间中，对应坐标空间中的点(x，y)的像素，其中x,y∈Z。p (x, y) represents the pixel corresponding to the point (x, y) in the coordinate space in the image space, where x, y∈Z.
dist(o1,o2)：当o1，o2为二维空间中的点时，dist(o1,o2)表示o1与o2间的欧几里得距离；当o1，o2分别为二维空间中的点和直线时，dist(o1,o2)表示从o1做与o2垂直的直线，垂线与o2的交点到o1的欧几里得距离；当o1,o2∈R，dist(o1,o2)＝|o1-o2|。dist(o 1 ,o 2 ): when o 1 and o 2 are points in two-dimensional space, dist(o 1 ,o 2 ) represents the Euclidean distance between o 1 and o 2 ; when o 1 , When o 2 is a point and a straight line in a two-dimensional space, dist(o 1 , o 2 ) represents the Euclidean distance from the point of intersection of the vertical line and o 2 to o 1 by making a straight line perpendicular to o 1 from o 1 ; When o 1 , o 2 ∈ R, dist(o 1 , o 2 )=|o 1 -o 2 |.
对于r，round(r)＝min{z∈Z+|dist(z,r)≥0}。For r, round(r)=min{z∈Z + |dist(z,r)≥0}.
偏移量：对于偏移量为 offset: for Offset is
序列化：将中的像素按索引值递增的顺序，向右逐个成行排列，当一行像素数为则在此行第一个像素正下方另起一行，并按上一行的排列方法继续排列像素，如此反复，直至中所有像素按索引值递增的顺序排列完毕，此时每行像素数为 可视作具有长宽且单位欧几里得距离为的二维空间。serialization: will The pixels in are arranged in rows one by one to the right in the order of increasing index value. When the number of pixels in a row is Then start another line directly below the first pixel in this line, and continue to arrange pixels according to the arrangement method of the previous line, and so on until All the pixels in are arranged in the order of increasing index value. At this time, the number of pixels in each row is can be regarded as having a long width and the unit Euclidean distance is two-dimensional space.
坐标空间(记作)：序列化并任选其中一个像素为原点的直角坐标系。对于该坐标系中具有单位欧几里得距离的两点(x1,y1)和(x2，y2)，由式给出的单位欧几里得距离值为该坐标系是具有无限长(y轴)宽(x轴)的二维空间。coordinate space (denoted as ):Serialization And choose one of the pixels as the origin of the Cartesian coordinate system. For two points (x 1 , y 1 ) and (x 2 , y 2 ) with unit Euclidean distance in this coordinate system, the formula The given unit Euclidean distance value is This coordinate system is a two-dimensional space with infinite length (y-axis) and width (x-axis).
平均中心：对于其中i＝1,2…n，n为BDSS所含像素总数；平均中心的坐标为 mean center: for Where i=1,2...n, n is the total number of pixels contained in the BDSS; average center The coordinates are
图像中心：即像素 Image center: i.e. pixel
坐标中心化：序列化设的原点为图像中心且x正半轴指向图像中心所在行的右侧，对于偏移量已知为di的像素与中坐标为 的点一一对应，(xi,yi)称为pi的中心坐标。当像素的中心坐标已知为(xi,yi)时，其偏移量为 索引值为 Coordinate Centering: Serialization Assume The origin of is the center of the image and the positive x-axis points to the right side of the row where the center of the image is located. For pixels whose offset is known as d i and The middle coordinates are The points correspond one-to-one, and ( xi , y ) is called the central coordinate of p i . when pixel When the center coordinate of is known as ( xi , y i ), its offset is The index value is
数字直线段(digital straight segment，DSS)：坐标中心化若像素和的坐标值分别为和ip,iq,jp,jq∈Z且ip＜iq，过p和q的直线记作且其斜率属于区间(0,1)，对于中直线x＝i∈R其中ip≤i≤iq，若直线和直线x＝i的交点坐标为(x,y),则存在坐标值为(round(x),round(y))的唯一像素，其与交点的欧几里得距离不大于在区间[ip,iq]内变化i值，可获得此类像素的集合，由p和q确定的数字直线段DSS(p,q)为：Digital straight segment (DSS): coordinate centering if pixel with The coordinate values are respectively with i p ,i q ,j p ,j q ∈ Z and i p <i q , the straight line passing through p and q is denoted as And its slope belongs to the interval (0,1), for In the straight line x=i∈R where i p ≤i≤i q , if the straight line The coordinates of the intersection point with the straight line x=i are (x, y), then there is a unique pixel whose coordinate value is (round(x), round(y)), and the Euclidean distance between it and the intersection point is not greater than Change the value of i within the interval [i p ,i q ] to obtain a set of such pixels, and the digital straight line segment DSS(p,q) determined by p and q is:
模糊数字直线段(blurred digital straight segment，BDSS)：坐标中心化若i,j,a,b,μ,ω∈Z,b≠0且gcd(a,b)＝1，则满足条件μ≤di-bj≤μ+ω的的合集称为模糊数字直线段BDSS(a,b,μ,ω)，即：Blurred digital straight segment (BDSS): coordinate centering If i,j,a,b,μ,ω∈Z,b≠0 and gcd(a,b)=1, the condition μ≤di-bj≤μ+ω is satisfied The collection of is called fuzzy digital straight line segment BDSS(a,b,μ,ω), namely:
在以上定义的基础上，本发明解决其技术问题所采用的技术方案是：On the basis of the above definitions, the technical solution adopted by the present invention to solve the technical problems is:
当模糊数字直线段BDSS识别算法完成对宽高的数字图像中模糊数字直线段BDSS的识别，模糊数字直线段BDSS以集合的形式保存为识别结果；When the fuzzy digital straight line segment BDSS recognition algorithm completes the width high The recognition of the fuzzy digital straight line segment BDSS in the digital image, the fuzzy digital straight line segment BDSS is saved as a recognition result in the form of a set;
根据识别结果给定的一个模糊数字直线段BDSS，计算其几何直线特性，使用斜率符合该几何直线特性的矩形边框标识给定像素集合。According to a fuzzy digital straight segment BDSS given by the recognition result, its geometric straight line characteristics are calculated, and a given set of pixels is identified by a rectangular border whose slope conforms to the geometric straight line characteristics.
具体包括以下步骤：Specifically include the following steps:
步骤1：计算坐标中心化图像空间中，对应坐标空间的中心坐标；Step 1: Calculate the corresponding coordinate space in the coordinate centered image space the center coordinates of
步骤2：根据模糊数字直线段BDSS中像素的中心坐标，计算平均中心，并使用线性回归计算模糊数字直线段BDSS的方向θ；Step 2: Calculate the average center according to the center coordinates of pixels in the blurred digital straight segment BDSS, and use linear regression to calculate the direction θ of the fuzzy digital straight segment BDSS;
步骤3：计算模糊数字直线段BDSS中像素的转换坐标：将坐标中心化对应的坐标空间的原点平移到平均中心，并将x轴正半轴旋转至与θ重合；Step 3: Calculate the transformation coordinates of the pixels in the blurred digital straight line segment BDSS: center the coordinates in the corresponding coordinate space The origin of is translated to the mean center, and the positive semi-axis of the x-axis is rotated to coincide with θ;
步骤4：根据模糊数字直线段BDSS中像素的转换坐标计算模糊数字直线段BDSS在转换坐标系中的四个边界点；Step 4: Calculate the fuzzy digital straight line segment BDSS in the transformed coordinate system according to the transformed coordinates of the pixels in the fuzzy digital straight line segment BDSS The four boundary points in ;
步骤5：生成平行于方向θ的边界线l1和l2，并计算其转换坐标，边界线l1和l2分别通过步骤4中的其中两个边界点；Step 5: Generate boundary lines l 1 and l 2 parallel to the direction θ, and calculate their transformation coordinates. The boundary lines l 1 and l 2 respectively pass through two of the boundary points in step 4;
步骤6：生成垂直于方向θ的边界线l3和l4，并计算其转换坐标，边界线l3和l4分别通过步骤4中的另外两个边界点，直线l1，l2，l3和l4相交形成矩形边框；Step 6: Generate the boundary lines l 3 and l 4 perpendicular to the direction θ, and calculate their transformation coordinates, the boundary lines l 3 and l 4 respectively pass through the other two boundary points in step 4, straight lines l 1 , l 2 , l 3 and l 4 intersect to form a rectangular frame;
步骤7：根据步骤5和6所生成的转换坐标，计算其中心坐标；Step 7: Calculate the center coordinates according to the transformation coordinates generated in steps 5 and 6;
步骤8：根据步骤7所生成的中心坐标，计算其偏移量并将偏移量保存为结果输出。Step 8: According to the center coordinates generated in step 7, calculate its offset and save the offset as the result output.
根据偏移量所确定像素的集合即为矩形边框，最后，可以使用像素修改方法，在识别BDSS的数字图像中更改矩形边框所含像素的值以标识模糊数字直线段BDSS。The set of pixels determined according to the offset is the rectangular frame. Finally, the pixel modification method can be used to change the value of the pixels contained in the rectangular frame in the digital image for identifying BDSS to identify the blurred digital straight line segment BDSS.
步骤1和2中，将识别模糊数字直线段BDSS的数字图像映射为坐标中心化对于任意像素p∈BDSS，若p的偏移量为d，则其中心坐标根据式(1)计算：In steps 1 and 2, the digital image for identifying the blurred digital straight line segment BDSS is mapped as Coordinate centering For any pixel p ∈ BDSS, if the offset of p is d, its center coordinates are calculated according to formula (1):
通过BDSS像素的中心坐标，计算BDSS的平均中心 Calculate the average center of the BDSS through the center coordinates of the BDSS pixels
根据线性回归理论，BDSS的方向θ∈[0，π)，由式(2)计算：According to the linear regression theory, the direction θ∈[0, π) of BDSS is calculated by formula (2):
其中i＝1,2…n，n为BDSS所含像素总数。 Where i=1,2...n, n is the total number of pixels contained in the BDSS.
步骤3中，对于其转换坐标根据式(3)计算：In step 3, for its transformed coordinates Calculate according to formula (3):
步骤4中，转换坐标系中的四个边界点通过比较转换坐标的所有横坐标x和纵坐标y得到。In step 4, transform the coordinate system The four boundary points in are obtained by comparing all abscissa x and ordinate y of transformed coordinates.
坐标空间的原点的转换坐标为 步骤7中，对于其中心坐标根据式(7)计算：coordinate space The transformed coordinates of the origin of are In step 7, for its center coordinates Calculate according to formula (7):
步骤8中，对于像素的偏移量d由式(8)计算：In step 8, for The pixel offset d is calculated by formula (8):
在中，由于原点位于BDSS的平均中心，BDSS像素转换坐标存在四个极值，即横坐标最大值x′lmax，横坐标最小值x′lmin，纵坐标最大值y′wmax和纵坐标最小值y′wmin；中存在着对应转换坐标(x′lmax,0)，(x′lmin,0)，(0,y′wmax)和(0,y′wmin)的像素，这些像素为步骤4所述的边界点；exist In , since the origin is located at the average center of BDSS, there are four extreme values in BDSS pixel conversion coordinates, namely, the maximum value x′ lmax of the abscissa, the minimum value x′ lmin of the abscissa, the maximum value y′ wmax of the ordinate and the minimum value y of the ordinate 'wmin; There are pixels corresponding to the transformation coordinates (x′ lmax ,0), (x′ lmin ,0), (0,y′ wmax ) and (0,y′ wmin ), these pixels are the boundary points described in step 4 ;
其中，表示矩形边框，矩形边框的定义由式(4)给出：in, Indicates a rectangular border, a rectangular border The definition of is given by formula (4):
其中m＝1，2，3，4，表示从p′(x′,y′)做与lm垂直的直线，垂线与lm的交点到p′(x′,y′)的欧几里得距离。in m=1, 2, 3, 4, Indicates the Euclidean distance from p′ ( x′, y′) to p′ (x′, y′) from the point of intersection of the vertical line and l m to p ′ (x′, y′).
矩形边框实际上由四条线段组成，式(4)可被替换为式(5)和式(6)，式(5)和式(6)分别为l1和l2上，l3和l5上像素的转换坐标：rectangular border Actually consists of four line segments, formula (4) can be replaced by formula (5) and formula (6), formula (5) and formula (6) are l 1 and l 2 , l 3 and l 5 Transformed coordinates of pixels:
本发明的有益效果是，本发明的一种用于标识数字图像中给定模糊数字直线段的方法，对给定数字直线段，模糊数字直线段等，由分散或连续像素组成的像素集合，计算其几何直线特性，使用斜率符合所计算特性的矩形边框标识给定像素集合的计算方法，具有如下优点：The beneficial effect of the present invention is, a kind of method of the present invention is used for identifying given fuzzy digital straight line segment in digital image, for given digital straight line segment, fuzzy digital straight line segment etc., the pixel collection that is made up of scattered or continuous pixel, Calculate its geometric straight line characteristics, and use the calculation method of identifying a given pixel set with a rectangular border whose slope conforms to the calculated characteristics, which has the following advantages:
(1)本发明方法在未大幅修改背景像素的前提下，清晰地标明了给定的模糊数字直线段。随着被标识数字图像分辨率的上升，本发明方法所修改像素的数量与图像全部像素之比快速下降，其视觉效果呈现为高分别率图像中，本发明标识的BDSS背景更容易观察；即使是小分辨率图像，本发明方法对图像背景的占用也较有限。(1) The method of the present invention clearly marks a given fuzzy digital straight line segment without greatly modifying the background pixels. Along with the rise of the resolution of the marked digital image, the ratio of the number of pixels modified by the method of the present invention to all pixels of the image drops rapidly, and its visual effect appears as a high-resolution image, and the BDSS background marked by the present invention is easier to observe; even It is a small-resolution image, and the method of the present invention occupies relatively limited image background.
(2)本发明方法实施简单，耗时与分辨率基本呈线性稳定增加，性能高效。(2) The method of the present invention is simple to implement, the time consumption and resolution increase substantially linearly and steadily, and the performance is efficient.
附图说明Description of drawings
下面结合附图和实施例对本发明进一步说明。The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
图1是目前常用的标识方法。(a)端点标识法，(b)彩色线段标识法，(c)斜率覆盖标识法。Figure 1 is the commonly used identification method at present. (a) endpoint identification method, (b) color line segment identification method, (c) slope coverage identification method.
图2是本发明的一种用于标识数字图像中给定模糊数字直线段的方法的原理图。(a)将原坐标系旋转到转换坐标系，(b)将转换坐标系旋转回原坐标系。FIG. 2 is a schematic diagram of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention. (a) Rotate the original coordinate system to the transformation coordinate system, (b) rotate the transformation coordinate system back to the original coordinate system.
图3是本发明的一种用于标识数字图像中给定模糊数字直线段的方法中Point类的数据结构。Fig. 3 is a data structure of the Point class in a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图4是本发明的一种用于标识数字图像中给定模糊数字直线段的方法的原理框图。Fig. 4 is a functional block diagram of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图5是本发明的一种用于标识数字图像中给定模糊数字直线段的方法中步骤2的原理框图。Fig. 5 is a functional block diagram of step 2 in a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图6是本发明的一种用于标识数字图像中给定模糊数字直线段的方法中步骤5的原理框图。Fig. 6 is a functional block diagram of step 5 in a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图7是本发明的一种用于标识数字图像中给定模糊数字直线段的方法中步骤6的原理框图。Fig. 7 is a functional block diagram of step 6 in a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图8是本发明标识128*128分辨率的图像中给定模糊数字直线段的结果图。(a)图像cameraman，(b)图像house，(c)图像lena，(d)图像puzzle。Fig. 8 is a result diagram of a given fuzzy digital straight line segment identified in an image with a resolution of 128*128 according to the present invention. (a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
图9是本发明标识256*256分辨率的图像中给定模糊数字直线段的结果图。(a)图像cameraman，(b)图像house，(c)图像lena，(d)图像puzzle。Fig. 9 is a result diagram of a given fuzzy digital straight line segment identified in an image with a resolution of 256*256 according to the present invention. (a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
图10是本发明标识512*512分辨率的图像中给定模糊数字直线段的结果图。(a)图像cameraman，(b)图像house，(c)图像lena，(d)图像puzzle。Fig. 10 is a result diagram of a given fuzzy digital straight line segment identified in an image with a resolution of 512*512 according to the present invention. (a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
图11是本发明的一种用于标识数字图像中给定模糊数字直线段的方法不同步骤在完成绘制一幅图中所有BDSS边框的平均耗时情况分析图。Fig. 11 is an analysis diagram of the average time-consuming situation of drawing all BDSS borders in a picture in different steps of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图12是本发明的一种用于标识数字图像中给定模糊数字直线段的方法不同步骤在完成绘制一幅图中最长的BDSS边框的耗时情况分析图。Fig. 12 is an analysis diagram of the time-consuming situation of drawing the longest BDSS border in a picture in different steps of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图13是本发明的一种用于标识数字图像中给定模糊数字直线段的方法不同步骤在完成绘制一幅图中最短的BDSS边框的耗时情况分析图。Fig. 13 is an analysis diagram of the time-consuming situation of drawing the shortest BDSS border in a picture in different steps of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention.
图14是本发明的一种用于标识数字图像中给定模糊数字直线段的方法总耗时随图像分辨率增加而变化的情况分析图。Fig. 14 is a situation analysis diagram of the total time consumption of a method for identifying a given fuzzy digital straight line segment in a digital image according to the present invention as the image resolution increases.
具体实施方式detailed description
现在结合附图对本发明作进一步详细的说明。The present invention is described in further detail now in conjunction with accompanying drawing.
当BDSS识别算法完成对宽高的数字图像中模糊数字直线段(BDSS)的识别，BDSS以集合的形式保存为识别结果。When the BDSS recognition algorithm completes the wide high The recognition of fuzzy digital straight line segment (BDSS) in the digital image, BDSS is saved as the recognition result in the form of a set.
对于任一BDSS，根据BDSS定义，其形态类似附图说明的图2(a)和图2(b)所示。图2(a)和图2(b)用方块代表的像素群分别展示了一个BDSS，图2(a)表示将原坐标系旋转到转换坐标系，图2(b)表示将转换坐标系旋转回原坐标系。本发明的目标是根据给定的一个BDSS，通过基于数学原理的计算，得到如图2(a)所示，由直线l1，l2，l3和l4相交形成的矩形边框在识别该BDSS的数字图像中，所对应像素的偏移量。每条直线通过一个边界点，且l1，l2与BDSS方向θ平行，l3，l4与θ垂直。For any BDSS, according to the definition of BDSS, its shape is similar to that shown in Figure 2(a) and Figure 2(b) in the accompanying drawings. Figure 2(a) and Figure 2(b) represent a BDSS by the pixel groups represented by squares, Figure 2(a) represents the rotation of the original coordinate system to the transformation coordinate system, and Figure 2(b) represents the rotation of the transformation coordinate system Return to the original coordinate system. The goal of the present invention is to obtain a rectangular border formed by the intersection of straight lines l 1 , l 2 , l 3 and l 4 as shown in Figure 2(a) in identifying the BDSS In the digital image of BDSS, the offset of the corresponding pixel. Each straight line passes through a boundary point, and l 1 , l 2 are parallel to BDSS direction θ, and l 3 , l 4 are perpendicular to θ.
根据偏移量所确定像素的集合称为矩形边框，可以使用本发明以外的各种像素修改方法，在识别BDSS的数字图像中更改矩形边框所含像素的值以标识BDSS。The set of pixels determined according to the offset is called a rectangular frame, and various pixel modification methods other than the present invention can be used to change the values of the pixels contained in the rectangular frame to identify the BDSS in the digital image for identifying BDSS.
对于给定的一个BDSS，将识别BDSS的数字图像映射为坐标中心化 中像素中心坐标对应的坐标系在图2中由相互垂直的箭头直线表示，标有x的箭头表示x轴正半轴。对于任意像素p∈BDSS，若p的偏移量为d，则其中心坐标根据式(1)计算：For a given BDSS, the digital image identifying the BDSS is mapped as Coordinate centering The coordinate system corresponding to the pixel center coordinates in In Fig. 2, it is represented by mutually perpendicular arrow lines, and the arrow marked with x represents the positive semi-axis of the x-axis. For any pixel p ∈ BDSS, if the offset of p is d, its center coordinates are calculated according to formula (1):
通过BDSS像素的中心坐标，可计算BDSS的平均中心图2中以实心小点和汉字标识了根据线性回归理论，BDSS的方向θ∈[0，π)由式(2)计算：Through the center coordinates of BDSS pixels, the average center of BDSS can be calculated In Figure 2, it is marked with solid dots and Chinese characters According to the linear regression theory, the direction θ∈[0, π) of BDSS is calculated by formula (2):
n为BDSS所含像素总数。n is the total number of pixels contained in the BDSS.
图2(a)中l1和l2平行，l1和l2与中x轴正半轴的较小夹角为θ。根据θ和BDSS像素的中心坐标虽可直接计算l1，l2，l3和l4参数及交点，但计算过程较复杂，且实施难度较大。In Figure 2(a), l 1 and l 2 are parallel, and l 1 and l 2 are with The smaller included angle of the positive semi-axis of the x-axis is θ. Although the l 1 , l 2 , l 3 , and l 4 parameters and intersection points can be directly calculated according to θ and the center coordinates of BDSS pixels, the calculation process is complicated and difficult to implement.
本发明规避了直接计算l1，l2，l3和l4交点的传统方法，通过平移和旋转坐标系并计算在转换后坐标系中BDSS的最大和最小坐标值，完成矩形边框的计算。平移是将的原点移动到BDSS的平均中心，旋转是逆时针旋转的x轴正半轴，并保证旋转后的x轴正半轴与原x轴正半轴的夹角为θ，通过这种平移和旋转获得的坐标系称为转换坐标系，记作图2(a)中，由相互垂直的箭头直线表示,标有x′的箭头表示x轴正半轴，标有y′的箭头表示y轴正半轴，弧形箭头虚线从原点指向原点表示将的原点平移至与的x轴正半轴的夹角为θ，旋转方向由弧形箭头实线表示。BDSS像素在中的坐标称为转换坐标。对于其转换坐标根据式(3)计算：The present invention avoids the traditional method of directly calculating the intersection of l 1 , l 2 , l 3 and l 4 , and completes the calculation of the rectangular frame by translating and rotating the coordinate system and calculating the maximum and minimum coordinate values of the BDSS in the converted coordinate system. Translation is to The origin moves to the mean center of the BDSS, and the rotation is counterclockwise The positive semi-axis of the x-axis, and ensure that the angle between the positive semi-axis of the x-axis after rotation and the positive semi-axis of the original x-axis is θ. The coordinate system obtained through this translation and rotation is called the transformation coordinate system, denoted as In Figure 2(a), Indicated by mutually perpendicular arrow lines, the arrow marked with x' indicates the positive semi-axis of the x-axis, the arrow marked with y' indicates the positive semi-axis of the y-axis, and the dotted line of the arc-shaped arrow starts from origin point origin express will The origin of is translated to and The included angle of the positive x-axis of the x-axis is θ, and the direction of rotation is indicated by the solid line of the arc-shaped arrow. BDSS pixels in The coordinates in are called transformed coordinates. for its transformed coordinates Calculate according to formula (3):
在中，由于原点位于BDSS的平均中心，BDSS像素转换坐标存在四个极值，即横坐标最大值x′lmax，横坐标最小值x′lmin，纵坐标最大值y′wmax和纵坐标最小值y′wmin。中存在着对应转换坐标(x′lmax,0)，(x′lmin,0)，(0,y′wmax)和(0,y′wmin)的像素，这些像素称为边界点，图2(a)中用实心小点和转换坐标标识了边界点。exist In , since the origin is located at the average center of BDSS, there are four extreme values in BDSS pixel conversion coordinates, namely, the maximum value x′ lmax of the abscissa, the minimum value x′ lmin of the abscissa, the maximum value y′ wmax of the ordinate and the minimum value y of the ordinate ' wmin . There are pixels corresponding to the transformation coordinates (x′ lmax ,0), (x′ lmin ,0), (0,y′ wmax ) and (0,y′ wmin ), these pixels are called boundary points, Figure 2( In a), the boundary points are marked with solid dots and transformed coordinates.
通过边界点，θ，l1与l2平行，l3与l4平行且l1与l3垂直的条件，对于图2(a)中所示的情况，矩形边框的定义由式(4)给出：Through the boundary point, θ, the condition that l 1 is parallel to l 2 , l 3 is parallel to l 4 and l 1 is perpendicular to l 3 , for Fig. 2(a) As shown in the case, the rectangular border The definition of is given by formula (4):
其中dm＝dist(p′(x′,y′),lm)，m＝1，2，3，4，dist(p′(x′,y′),lm)表示从p′(x′,y′)做与lm垂直的直线，垂线与lm的交点到p′(x′,y′)的欧几里得距离。Where d m =dist(p′ (x′,y′) ,l m ), m=1,2,3,4, dist(p′ (x′,y′) ,l m ) means from p′ ( x′, y′) make a straight line perpendicular to l m , and the Euclidean distance from the intersection of the vertical line and l m to p′ (x′, y′) .
对于图2(a)中所示的情况，实际上由四条线段组成，式(4)可被式(5)和式(6)替代：For Figure 2(a) In the case shown in the Actually consists of four line segments, formula (4) can be replaced by formula (5) and formula (6):
式(5)和式(6)分别给出了l1和l2上，l3和l4上像素的转换坐标。通过像素的转换坐标，可以计算其像素偏移量，通过偏移量可更改识别BDSS的数字图像中对应像素的值。计算偏移量的过程依次包括根据转换坐标计算中心坐标和根据中心坐标计算偏移量。Formula (5) and formula (6) respectively give l 1 and l 2 up, l 3 and l 4 up Transformed coordinates of pixels. pass The converted coordinates of the pixel can calculate its pixel offset, and the value of the corresponding pixel in the digital image for identifying the BDSS can be changed through the offset. The process of calculating the offset sequentially includes calculating the center coordinates based on the transformed coordinates and calculating the offset based on the center coordinates.
根据转换坐标计算中心坐标，相当于通过将的原点平移回的原定，并顺时针旋转中x轴正半轴至的x轴正半轴，将还原为附图说明的图2(b)中，由相互垂直的箭头直线表示，标有x′的箭头表示x轴正半轴，标有y′的箭头表示y轴正半轴，由相互垂直的箭头直线表示,标有x的箭头表示x轴正半轴，从原点指向原点的弧形箭头虚线表示原点平移，从的x轴正半轴指向的x轴正半轴的弧形箭头实线表示逆时针旋转。还原过程需要原点的转换坐标，其转换坐标为对于 其中心坐标根据式(7)计算：Calculate the center coordinates according to the transformation coordinates, which is equivalent to passing shift back to the origin of of the original, and rotate clockwise The positive semi-axis of the x-axis to The positive x-axis of the x-axis will be revert to In Figure 2(b) of the accompanying drawings, Indicated by mutually perpendicular arrow lines, the arrow marked with x' indicates the positive semi-axis of the x-axis, and the arrow marked with y' indicates the positive semi-axis of the y-axis, It is represented by mutually perpendicular arrow lines, and the arrow marked with x represents the positive semi-axis of the x-axis, from origin direction The dotted line of the arc-shaped arrow at the origin represents the translation of the origin, from The positive x-axis of the x-axis points to The solid line of the arc-shaped arrow of the positive x-axis indicates counterclockwise rotation. The restore process requires The transformed coordinates of the origin, whose transformed coordinates are for its center coordinates Calculate according to formula (7):
根据像素的中心坐标，对于像素的偏移量d由式(8)计算：according to The coordinates of the center of the pixel, for The pixel offset d is calculated by formula (8):
通过像素的偏移量，即可在识别BDSS所用的数字图像中更改像素的值，从而完成标识模糊数字直线段BDSS的目的。pass The pixel offset that can be changed in the digital image used to identify the BDSS The value of the pixel, so as to complete the purpose of identifying the blurred digital straight line segment BDSS.
具体地，本发明计算方法的实施方式由编写计算机程序完成。Specifically, the implementation of the calculation method of the present invention is accomplished by writing computer programs.
程序输入包括：一个BDSS，识别BDSS所使用的数字图像以及其宽和高其中BDSS的像素以附图说中图3所示数据结构表示。图3展示了Point类，其属性包括用于存储偏移量的Deviation，分别用于存储中心坐标横坐标x的属性XtoImageCenter和纵坐标y的属性YtoImageCenter，分别用于存储转换坐标横坐标x的属性XtoCollectionCenter和纵坐标y的属性YtoCollectionCenter。属性Deviation的数据类型为整数，用integer表示，其他Point类属性的数据类型为双精度浮点数，用double表示。The program input includes: a BDSS, the digital image used to identify the BDSS and its width and high The pixels of the BDSS are represented by the data structure shown in Figure 3 in the accompanying drawings. Figure 3 shows the Point class, whose properties include Deviation for storing offsets, XtoImageCenter for storing the center coordinate abscissa x, and YtoImageCenter for the y coordinate y, and YtoImageCenter for storing the transformation coordinate abscissa x The property YtoCollectionCenter of XtoCollectionCenter and ordinate y. The data type of the property Deviation is an integer, represented by integer, and the data type of other Point class properties is a double-precision floating point number, represented by double.
程序输出为根据输入BDSS计算的，由式(4)定义的像素集合。The output of the program is the set of pixels defined by Equation (4), calculated from the input BDSS.
程序包括八个步骤，见附图说明的图4。现依次介绍八个步骤。The procedure consists of eight steps, see Figure 4 of the legend to the drawings. The eight steps are described in sequence.
步骤1：计算坐标中心化图像空间中，对应坐标空间的Point对象的中心坐标。Step 1: Calculate the center coordinates of the Point object corresponding to the coordinate space in the coordinate centered image space.
映射数字图像至图像空间根据中像素的索引值，输入的和可计算得到每个Point对象的偏移量并坐标中心化对于偏移量为d的Point对象，其中心坐标根据式(1)计算，中心坐标的横坐标x与纵坐标y分别以属性XtoImageCenter和YtoImageCenter保存。Mapping digital images to image space according to The index value of the pixel in the input with The offset of each Point object can be calculated and the coordinates can be centered For a Point object with an offset of d, its center coordinates are calculated according to formula (1), and the abscissa x and ordinate y of the center coordinates are saved with attributes XtoImageCenter and YtoImageCenter respectively.
步骤2：根据BDSS中Point对象坐标，计算平均中心，并使用线性回归计算其方向θ。Step 2: According to the coordinates of the Point object in BDSS, calculate the average center, and use linear regression to calculate its direction θ.
根据步骤1计算的BDSS中各Point对象中心坐标，计算BDSS的平均中心基于线性回归理论，根据式(2)计算BDSS的方向θ。具体流程见附图说明的图6。According to the center coordinates of each Point object in BDSS calculated in step 1, calculate the average center of BDSS Based on the linear regression theory, the direction θ of the BDSS is calculated according to formula (2). The specific process is shown in Figure 6 of the accompanying drawings.
步骤3：将坐标中心化对应的坐标空间的原点平移到平均中心，并将x轴正半轴旋转至与θ重合，计算BDSS中Point对象的转换坐标。Step 3: Translate the origin of the coordinate space corresponding to the coordinate centering to the average center, and rotate the positive semi-axis of the x-axis to coincide with θ, and calculate the transformation coordinates of the Point object in BDSS.
Point对象的转换坐标根据式(3)计算，转换坐标的横坐标x与纵坐标y分别以属性XtoCollectionCenter和YtoCollectionCenter保存。The conversion coordinates of the Point object are calculated according to formula (3), and the abscissa x and y coordinates of the conversion coordinates are saved with the attributes XtoCollectionCenter and YtoCollectionCenter respectively.
步骤4：根据BDSS中Point对象的转换坐标计算边界点。Step 4: Calculate the boundary points according to the transformed coordinates of the Point object in BDSS.
中BDSS坐标极值是通过比较BDSS中所有Point对象转换坐标的XtoCollectionCenter属性值和YtoCollectionCenter属性值得到。根据比较得到的极值，生成类似附图说明的图2中边界点(x′lmax,0)，(x′lmax,0)，(0,y′wmax)和(0,y′wmax)的Point对象。这些边界点是完成步骤5和步骤6计算的基础。 The extreme value of BDSS coordinates is obtained by comparing the XtoCollectionCenter attribute value and YtoCollectionCenter attribute value of all Point objects in BDSS to convert coordinates. According to the extremum obtained by comparison, generate the boundary points (x′ lmax ,0), (x′ lmax ,0), (0,y′ wmax ) and (0,y′ wmax ) in Figure 2 similar to the illustration Point object. These boundary points are the basis for completing the calculations in steps 5 and 6.
步骤5：生成平行于方向θ的边界线的Point对象，并计算其转换坐标。Step 5: Generate a Point object of the boundary line parallel to the direction θ, and calculate its transformed coordinates.
根据式(5)生成对应图2中l1和l2的Point对象，具体流程见附图说明的图6。边界点的坐标分别用变量Xstart，Xend，Ystart和Yend保存。以附图说明的图2中的边界点(x′lmax,0)，(x′lmin,0)，(0,y′wmax)和(0,y′wmin)为例，Xstart，Xend，Ystart和Yend的值分别为x′lmin，x′lmax，y′wmin和y′wmax，即横坐标x的最小值和最大值，纵坐标y的最小值和最大值。在保持Ystart和Yend不变且Xstar不大于Xend的前提下，以Xstart为转换坐标横坐标x的初值并每次循环Xstart自增1的方式，循环生成转换坐标分别为(Xstart,Ystart)和(Xend,Ystart)的两个Point对象。将循环产生的所有Point对象保存在矩形边框中。Point objects corresponding to l1 and l2 in Figure 2 are generated according to formula (5), and the specific process is shown in Figure 6 of the accompanying drawings. The coordinates of the boundary points are saved with the variables X start , X end , Y start and Y end respectively. Taking the boundary points (x' lmax , 0) in Figure 2 illustrated in the drawings, (x' lmin , 0), (0, y' wmax ) and (0, y' wmin ) as an example, X start , X end , the values of Y start and Y end are respectively x′ lmin , x′ lmax , y′ wmin and y′ wmax , that is, the minimum and maximum values of the abscissa x, and the minimum and maximum values of the y coordinate. On the premise of keeping Y start and Y end unchanged and X star not greater than X end , take X start as the initial value of the abscissa x of the conversion coordinate and increase by 1 each time X start is cycled, and the cyclically generated conversion coordinates are respectively Two Point objects of (X start , Y start ) and (X end , Y start ). Save all the Point objects generated by the loop in the bounding rectangle.
步骤6：生成垂直于方向θ的边界线的Point对象，并计算其转换坐标。Step 6: Generate a Point object of the boundary line perpendicular to the direction θ, and calculate its transformed coordinates.
根据式(6)生成对应图2中l3和l4的Point对象，具体流程见附图说明的图7。使用边界点横坐标x的最小值重新初始化变量Xstart，在保持Xstart和Xend不变且Ystart不大于Yend的前提下，以Ystart为转换坐标横坐标y的初值并每次循环Ystart自增1的方式，循环生成转换坐标分别为(Xstart,Ystart)和(Xstart,Yend)的两个Point对象。将循环产生的所有Point对象保存在矩形边框中。Point objects corresponding to l3 and l4 in Figure 2 are generated according to formula (6), and the specific process is shown in Figure 7 of the accompanying drawings. Use the minimum value of the abscissa x of the boundary point to reinitialize the variable X start . On the premise that X start and X end are kept unchanged and Y start is not greater than Y end , use Y start as the initial value of the abscissa y of the transformation coordinate and repeat each time By looping Y start and incrementing by 1, the loop generates two Point objects whose conversion coordinates are (X start , Y start ) and (X start , Y end ), respectively. Save all the Point objects generated by the loop in the bounding rectangle.
步骤7：根据所生成Point对象的转换坐标，计算其中心坐标。Step 7: According to the converted coordinates of the generated Point object, calculate its center coordinates.
对于矩形边框中所有Point对象，其中心坐标的计算根据式(7)完成。For all Point objects in the rectangular frame, the calculation of its center coordinates is completed according to formula (7).
步骤8：根据所生成Point对象的中心坐标，计算其偏移量并将偏移量保存为结果输出。Step 8: According to the center coordinates of the generated Point object, calculate its offset and save the offset as the result output.
对于矩形边框中所有Point对象，其偏移量的计算根据式(8)完成。For all Point objects in the rectangular border, the calculation of its offset is completed according to formula (8).
根据输出结果，可以使用像素修改方法，在识别BDSS的数字图像中更改矩形边框所含像素的值以标识模糊数字直线段BDSS。According to the output result, the pixel modification method can be used to change the value of the pixels contained in the rectangular frame in the digital image for identifying BDSS to identify the blurred digital straight segment BDSS.
以下分为两部介绍本发明的有益效果，第一部分为本发明方法的直观视觉效果；第二部分为计算效率分析：The following is divided into two parts to introduce the beneficial effects of the present invention, the first part is the intuitive visual effect of the method of the present invention; the second part is the calculation efficiency analysis:
1、视觉效果：1. Visual effects:
图8至图10展示了应用本发明方法标识给定BDSS的结果图(图中矩形边框均以1像素宽的黑线，白线，黑线的顺序叠加而成的线段表示)。图8至图10中，数字图像的分辨率依次为128×128，256×256和512×512。注意图像分辨率改动后，输入的BDSS也会变化，本发明方法的输出会随之变化，并不是放大相同的结果。图8至图10用于说明随分辨率提高，算法污染的背景像素相应减少，视觉上的感受是边框遮挡的背景变少，且更容易分辨相互叠加的BDSS。Figures 8 to 10 show the results of applying the method of the present invention to identify a given BDSS (the rectangular borders in the figure are represented by a black line, a white line, and a line segment superimposed in the order of a black line with a width of 1 pixel). In Fig. 8 to Fig. 10, the resolutions of digital images are 128×128, 256×256 and 512×512 in sequence. Note that after the image resolution is changed, the input BDSS will also change, and the output of the method of the present invention will change accordingly, which is not the same result of zooming in. Figures 8 to 10 are used to illustrate that as the resolution increases, the background pixels polluted by the algorithm decrease accordingly. The visual experience is that the background occluded by the border becomes less, and it is easier to distinguish BDSSs that overlap each other.
比较图8至图10所示结果与图1所示结果，很容易得出本发明方法在未大幅修改背景像素的前提下，清晰地标明了给定的BDSS，可以试想若采用图1(c)展示的算法标识分辨率较低的图像，例如图8，则被本发明方法标识的BDSS会以单色填充的矩形表示，从而导致人眼无法辨识BDSS背景区域。随着被标识数字图像分辨率的上升，本发明方法所修改像素的数量与图像全部像素之比快速下降，其视觉效果呈现为高分别率图像中，本发明标识的BDSS背景更容易观察，例如图10，即使是小分辨率图像，本发明方法对图像背景的占用也较有限，例如图8。Comparing the results shown in Figures 8 to 10 with the results shown in Figure 1, it is easy to conclude that the method of the present invention clearly marks a given BDSS without greatly modifying the background pixels. ) shows an image with a lower resolution, such as in Figure 8, the BDSS identified by the method of the present invention will be represented by a monochrome filled rectangle, resulting in the human eye being unable to identify the BDSS background area. As the resolution of the marked digital image increases, the ratio of the number of pixels modified by the method of the present invention to all pixels of the image decreases rapidly, and its visual effect appears as a high-resolution image. The BDSS background marked by the present invention is easier to observe, for example In Fig. 10 , even for a small-resolution image, the method of the present invention occupies a relatively limited image background, such as Fig. 8 .
2、计算效率：2. Calculation efficiency:
附图说明的图11至图13的柱状图展示了具体实施方式所述的八个步骤消耗的时间，注意图11至图13纵轴单位为CPU时钟周期数的对数。图11至图13横轴首先按图像cameraman，lena，house和puzzle的顺序从左到右排列，对于每一幅图，再按分别率128×128，256×256和512×512的顺序从左到右排列。以图11为例，横轴说明文字“图像cameraman”上方从左到右的立柱分别对应分别率为128×128，256×256和512×512的耗时。每个立柱里不同颜色的区域表示不同步骤的耗时，区域越大，则该步骤耗时越多，立柱越高，则总耗时越多。图11至图13中的同一立柱由下至上的各个方块依次表示具体实施方式所述的步骤1至步骤8的每个步骤所消耗的时间。图11展示了平均耗时，即对一幅数字图像中的所有给定BDSS的标识各步骤及总耗时平均值。图12展示了最大耗时，即对一幅数字图像中给定的具有最多Point对象的BDSS标识中各步骤及总耗时。图13展示了最小耗时，即对一幅数字图像中给定的具有最少Point对象的BDSS标识中各步骤及总耗时。图11-13是显示不同步骤在不同情况下的耗时分布，用于说明算法核心步骤5，6，7耗时较少。The histograms in Figures 11 to 13 in the description of the drawings show the time consumed by the eight steps described in the specific embodiment, and note that the unit of the vertical axis in Figures 11 to 13 is the logarithm of the number of CPU clock cycles. The horizontal axis of Figure 11 to Figure 13 is first arranged from left to right in the order of cameraman, lena, house and puzzle, and for each picture, in the order of resolution 128×128, 256×256 and 512×512 from left Arrange to the right. Taking Figure 11 as an example, the columns above the explanatory text "image cameraman" on the horizontal axis from left to right correspond to the time-consuming resolutions of 128×128, 256×256 and 512×512 respectively. The areas of different colors in each column represent the time consumption of different steps. The larger the area, the more time-consuming the step is, and the higher the column, the more time-consuming the total. The blocks in the same column in Fig. 11 to Fig. 13 represent the time consumed by each step from step 1 to step 8 described in the detailed description in sequence from bottom to top. Fig. 11 shows the average time consumption, that is, the average time consumption of each step and the total time consumption of all given BDSSs in a digital image. Fig. 12 shows the maximum time consumption, that is, each step and the total time consumption in the identification of the given BDSS with the most Point objects in a digital image. Figure 13 shows the minimum time consumption, that is, each step and the total time consumption in the identification of a given BDSS with the least Point objects in a digital image. Figure 11-13 shows the time-consuming distribution of different steps in different situations, which is used to illustrate that the core steps 5, 6, and 7 of the algorithm are less time-consuming.
通过观察图11至图13，步骤8在多个立柱中所占区域最多，即该步骤最耗时。这是由于步骤8计算复杂度较高，且要将矩形边框转换成符合后续处理格式而引起的。对于每个Point对象，根据式(7)步骤7要完成4次乘法和6次加法运算，而步骤8需完成1次乘法，2次除法，3次加法及2次取整运算虽然步骤8运算次数稍小，但其涉及的多是除法和取整此类较复杂的运算。用于计算矩形边框的步骤5和步骤6在图11至图13普遍耗时少于步骤3和步骤4，特别是当BDSS包含较少Point对象时，即图13所示。这从效率反应了本发明方法实施简单的特点，因为步骤5与步骤6是产生矩形边框的主要步骤，由于其实施过程未涉及复杂运算或数据结构，与其他步骤，特别是步骤7与步骤8相比，步骤5与步骤6耗时较少。By observing Fig. 11 to Fig. 13, step 8 occupies the largest area among the columns, that is, this step is the most time-consuming. This is due to the high computational complexity of step 8, and the need to convert the rectangular frame into a format suitable for subsequent processing. For each Point object, according to formula (7), step 7 needs to complete 4 multiplications and 6 additions, and step 8 needs to complete 1 multiplication, 2 divisions, 3 additions and 2 rounding operations Although the number of operations in step 8 is slightly smaller, it mostly involves more complex operations such as division and rounding. Steps 5 and 6 for calculating the rectangle border generally take less time than steps 3 and 4 in Figures 11 to 13, especially when the BDSS contains fewer Point objects, as shown in Figure 13. This has reflected the feature that the inventive method implements simply from efficiency, because step 5 and step 6 are the main steps of producing rectangular frame, because its implementation process does not involve complex calculation or data structure, and other steps, particularly step 7 and step 8 Compared with steps 5 and 6, it takes less time.
图14是总耗时随图像分辨率增加而变化的情况分析图，总耗时就是处理一幅图像中所有BDSS耗时的总和。图14展示了从分别率128×128到512×512，每间隔32×32的本发明计算方法总耗时情况，注意图14的横轴只显示了图像的一维大小。按上述分别率顺序，即128，160，192……，图像cameraman，lena，house和puzzle不同分辨率版本的耗时以实心圆点标注在图14中，对于相同内容但不同分别率图像的圆点用具有相同线段形状的直线段连接，具体的线段形状说明详见图14。图14中的四条连线基本呈线性，即耗时与分辨率基本成正比，其中典型的是图像cameraman的耗时分布，图像lena和图像puzzle的耗时分布虽然起伏很大，但从线性回归的角度分析，其分布与理想直线的差异较小。Figure 14 is an analysis diagram of the change of the total time consumption with the increase of the image resolution, and the total time consumption is the sum of the time consumption for processing all BDSS in an image. Figure 14 shows the total time consumption of the calculation method of the present invention for each interval of 32×32 from resolution 128×128 to 512×512, note that the horizontal axis of Figure 14 only shows the one-dimensional size of the image. According to the above resolution order, that is, 128, 160, 192..., the time consumption of different resolution versions of the images cameraman, lena, house and puzzle is marked in Figure 14 with solid dots, for the circles of images with the same content but different resolutions Points are connected by straight line segments with the same line segment shape, and the specific line segment shape description is shown in Figure 14. The four lines in Figure 14 are basically linear, that is, the time consumption is basically proportional to the resolution. The typical one is the time consumption distribution of the image cameraman, although the time consumption distribution of the image lena and the image puzzle fluctuates greatly, but from the linear regression Angular analysis of , the distribution of which differs little from the ideal straight line.
图14中，图像house的耗时分布随分辨率增长，出现下降的趋势，但由于其它图像均未出现这种趋势，house的情况仅是个别现象。综上，通过图14的耗时分布，未发现本发明方法随分辨率增加，明显呈现耗时非线性快速增加的情况，即，本发明方法耗时与分辨率基本呈线性稳定增加。这说明本发明计算方法的性能较为高效。In Figure 14, the time-consuming distribution of the image house shows a downward trend as the resolution increases, but since this trend does not appear in other images, the case of house is only an isolated phenomenon. To sum up, from the time-consuming distribution in Figure 14, it is not found that the method of the present invention shows a nonlinear and rapid increase in time-consuming as the resolution increases, that is, the time-consuming and resolution of the method of the present invention basically increase linearly and stably. This shows that the performance of the calculation method of the present invention is more efficient.
Claims (7)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510137710.3A CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510137710.3A CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104751177A CN104751177A (en) | 2015-07-01 |
| CN104751177B true CN104751177B (en) | 2018-01-12 |
Family
ID=53590830
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510137710.3A Active CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104751177B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513044B (en) * | 2015-11-20 | 2018-07-17 | 常州大学 | A kind of digital direct straight line segments recognition method based on statistical measures linear feature |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4901365A (en) * | 1988-12-19 | 1990-02-13 | Ncr Corporation | Method of searching binary images to find search regions in which straight lines may be found |
| CN102156884A (en) * | 2011-04-25 | 2011-08-17 | 中国科学院自动化研究所 | Straight segment detecting and extracting method |
| CN102819743A (en) * | 2012-08-14 | 2012-12-12 | 常州大学 | Detection method for quickly identifying straight-line segments in digital image |
| CN103150741A (en) * | 2012-11-30 | 2013-06-12 | 常州大学 | Method for rapidly skeletonizing graph of binary digital image |
-
2015
- 2015-03-26 CN CN201510137710.3A patent/CN104751177B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4901365A (en) * | 1988-12-19 | 1990-02-13 | Ncr Corporation | Method of searching binary images to find search regions in which straight lines may be found |
| CN102156884A (en) * | 2011-04-25 | 2011-08-17 | 中国科学院自动化研究所 | Straight segment detecting and extracting method |
| CN102819743A (en) * | 2012-08-14 | 2012-12-12 | 常州大学 | Detection method for quickly identifying straight-line segments in digital image |
| CN103150741A (en) * | 2012-11-30 | 2013-06-12 | 常州大学 | Method for rapidly skeletonizing graph of binary digital image |
Non-Patent Citations (1)
| Title |
|---|
| 一种新的不规则直线段识别方法;李沛等;《光学技术》;20100315;第295-301页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104751177A (en) | 2015-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Symmetrical feature propagation network for hyperspectral image super-resolution | |
| Heinrich et al. | State of the Art of Parallel Coordinates. | |
| CN105513110A (en) | GPU-based method for drawing linetype map symbol with gradually-varied width | |
| CN112100795B (en) | Computer aided design drawing comparison method and device | |
| Hu et al. | Efficient and automatic plane detection approach for 3-D rock mass point clouds | |
| Gustavson et al. | Anti-aliased Euclidean distance transform | |
| CN102270236A (en) | Rasterized geographic information system (GIS)-based spatial relationship judging method and system | |
| Shen et al. | A new approach to simplifying polygonal and linear features using superpixel segmentation | |
| US20240095888A1 (en) | Systems, methods, and devices for image processing | |
| DE112021000927T5 (en) | Systems, methods and devices for image processing | |
| CN111105431A (en) | Method and device for marking image | |
| CN104657999A (en) | Digital image correlation method based on kernel function | |
| Chen et al. | Extracting skeleton lines from building footprints by integration of vector and raster data | |
| CN116721230A (en) | Method, device, equipment and storage medium for constructing three-dimensional live-action model | |
| CN104751177B (en) | A kind of method for being used in reference numbers image give fuzzy digit straightway | |
| CN113126944A (en) | Depth map display method, display device, electronic device, and storage medium | |
| CN104574396B (en) | Straight line detection method and system | |
| CN107146215A (en) | A Saliency Detection Method Based on Color Histogram and Convex Hull | |
| Zhao et al. | DSC-Net: learning discriminative spatial contextual features for semantic segmentation of large-scale ancient architecture point clouds | |
| Ullah et al. | Experimental and numerical research of paved microcrack using histogram equalization for detection and segmentation | |
| CN113256645B (en) | Color image segmentation method based on improved density clustering | |
| Forsythe et al. | Convex constrained meshes for superpixel segmentations of images | |
| CN104156696B (en) | Bi-directional-image-based construction method for quick local changeless feature descriptor | |
| Kurlin et al. | A persistence-based approach to automatic detection of line segments in images | |
| Burch et al. | Partial link drawings for nodes, links, and regions of interest |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230414 Address after: Room 306, Block C, Tianrun Technology Building, Changzhou Science and Education City, No. 18-3 Changwu Middle Road, Wujin District, Changzhou City, Jiangsu Province, 213000 Patentee after: Changzhou Jingxun Micro Information Technology Co.,Ltd. Address before: Gehu Lake Road Wujin District 213164 Jiangsu city of Changzhou province No. 1 Patentee before: CHANGZHOU University |
|
| TR01 | Transfer of patent right |