CN103150735A - Gray level difference averaging-based image edge detection method - Google Patents

Abstract

The invention discloses a gray level difference averaging-based image edge detection method, which mainly comprises the following steps of: (1) starting reading an image; (2) setting a gray level threshold value; (3) converting the image into a gray level image; (4) filtering the image; (5) reading an image address; (6) calculating an absolute average value of a gray level difference; (7) judging whether the absolute average value of the gray level difference is greater than the threshold value or not, setting a gray level value to be 255 if the absolute average value of the gray level difference is greater than the threshold value, entering a step (8), and setting the gray level value to be 0 if the absolute average value of the gray level difference is not greater than the threshold value, and entering the step (8); and (8) judging whether the processing of the image is finished or not, performing edge extraction if the processing of the image is finished, and outputting an image edge, and adding 1 to the image address if the processing of the image is not finished, and returning to the step (5). The method is easy to understand and short in operation time, and an algorithm is simple; differential operation is not involved, the addition of noise in an image enhancement process is avoided, and a detection result under a proper threshold value is better.

Description

The average method for detecting image edge of a kind of intensity-based difference

Technical field

The present invention relates to the average method for detecting image edge of a kind of intensity-based difference, belong to the image processing and analysis field.

Background technology

Image Edge-Detection is one of most basic content in image processing and analysis, is also still not obtain so far the class problem that satisfactorily resolves.So-called edge refers to that its surrounding pixel gray scale has the set of those pixels of step variation or roof variation.The edge is present between object and background widely, between object and object, between primitive and primitive.To be that gray scale is discontinuous cause in its existence.The edge of image has comprised the features such as the position, profile of image, is one of essential characteristic of image, is widely used in the image processing and analysis field.Therefore, the detection method of image border is the study hotspot in the image processing and analysis technology always.

Yet the theory of the relevant rim detection of delivering so far and method remain in many weak points, such as being difficult to reach gratifying effect at aspects such as accuracy of detection, denoising and operation times.

The level of the classical detected edge image of roberts operator and vertical edge are more clear, but the discontinuous situation in edge easily appears in oblique line directions; Owing to not adopting the related measures such as smothing filtering, so anti-noise ability is relatively poor.

Rim detection effect and the sobel of Prewitt operator are close.Because all noise is had smoothing effect, so noiseproof feature is better; But detected edge is thicker, and bearing accuracy is not high, the situation of missing edges easily occurs.

The advantage that adopts Laplace operator is that the gained image border is thinner, and bearing accuracy is higher.But this algorithm is very responsive to noise, and Laplace operator affects meeting greater than the impact of its edge to noise sometimes.This is because Laplace operator adopts second differential, can doublely add like this impact of very noisy.

Log operator noiseproof feature is better, and this is because this algorithm adopts Gaussian filter to carry out filtering to image, effectively suppressed the noise in the original image, but monolateral response is good not, isolated point and pseudo-edge easily occur.

If only consider the rim detection effect, the canny operator as optimum operator, is obviously best selection; But when processing comparatively complicated picture, the calculated amount size becomes the factor of can not ignore of weighing the edge detection operator quality.In above-mentioned each classical operators, the calculated amount of canny operator is maximum.

Chinese patent (application number: 201110079661.4, patent name: based on the extracting method in the range gating image effective information district of average gray) described definite background area average gray value method, be that morphologic method is adopted in the extraction of largest connected background area; And the present invention need not to relate to morphology; The described priori of utilizing is calculated initial threshold, utilizes this initial threshold original image to be carried out the extraction in effective information district, calculates the average gray value at this edge, block of information according to the marginal portion, block of information of obtaining, and its threshold value adjustment need be carried out iterative operation; And the present invention need not iteration, and whole image is carried out single pass, and response is the marginal point of image greater than the point of threshold value, only by threshold values T just with Edge extraction out.

Summary of the invention

Purpose of the present invention is exactly to provide a kind of intensity-based difference average method for detecting image edge in order to address the above problem, it have advantages of the detected image edge and profile simple and effective.

To achieve these goals, the present invention adopts following technical scheme:

The average method for detecting image edge of a kind of intensity-based difference mainly comprises the steps:

Step (1): beginning, reading images;

Step (2): set gray threshold;

Step (3): image transitions is gray level image;

Step (4): image filtering;

Step (5): reading images address;

Step (6): calculate the gray scale difference value absolute average;

Step (7): whether judging the gray scale difference value absolute average greater than threshold value, is 255 if just set gray-scale value, and enters step (8); Be 0 if not just setting gray-scale value, enter step (8);

Step (8): judge whether image is handled, if just carry out edge extracting, the output image edge; Return to step (5) after just image address being added one.

The figure filtering of described step (4) is as follows: with Gaussian filter and median filter, image is carried out smothing filtering respectively, effectively suppress salt-pepper noise and Gaussian noise in image.

The concrete steps of the calculating gray scale difference value mean value of described step (6) are:

Absolute mean is asked in image pointwise after smothing filtering.Suppose g ₀Be the point of considering, select 3 * 3 neighborhoods, computing formula is as follows:

G _i＝|I(g ₀)-I(g _i)|

$G=\frac{\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{G}_i}{8}$

g1	g8	g7
			g2	g0	g6
g3	g4	g5

In formula, i is pixel, and span is 0～8; Gi represents the absolute difference of g0 point and gi point gray scale, I(gi) is the gray-scale value of pixel gi.G is at g ₀The response at place, this value representation g ₀The mean difference of the gray scale at some place and neighborhood territory pixel gray scale around it responds the character that G has embodied pixel.

The threshold value of described step (7) is chosen: according to the statistical property of image absolute mean matrix, carry out the threshold value of rim detection and choose

$\mathrm{avr}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\left|G(i,j)\right|}{M\×N}$

$\mathrm{std}=\sqrt{\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{(G(i,j)-\mathrm{avr})}^2}{M\×N}}$

T＝a×avr+b×std(a=1,b=0.5)

Wherein T represents selected threshold value, and M * N is picture size, and i, j are pixel, and span is respectively 1～M and 1～N; G (i, j) is the Grad that point (i, j) is located, and std and avr represent respectively standard deviation and the average of gradient matrix, and a, b are empirical constant, general a=1, b=0.5.When a, b get less value, can obtain more edge details; When a, b get larger value, can reduce the noise in edge image.

The edge extracting of described step (8):

Whole image is carried out single pass, and response is the marginal point of image greater than the point of threshold value, only by threshold values T just with Edge extraction out.

Beneficial effect of the present invention:

1, the edge of image has comprised the feature such as position, profile of image, is one of essential characteristic of image, can be widely used in biomedical engineering field;

2, can find out by MATLAB emulation, the method detected image border continuity is better, and pseudo-edge and isolated point less;

3, easy to understand, algorithm is simple, and operation time is short;

4, differentiate owing to not relating to, thus can not increase noise in the process of figure image intensifying, so testing result is better under suitable threshold value.

Description of drawings

Fig. 1 is main process flow diagram of the present invention;

Fig. 2 is the original image of embodiments of the invention;

Fig. 3 is the image after the processing of embodiments of the invention.

Embodiment

The invention will be further described below in conjunction with accompanying drawing and embodiment.

As shown in Figure 1, the average method for detecting image edge of a kind of intensity-based difference mainly comprises the steps:

Step (1): beginning, reading images;

Step (2): set gray threshold;

Step (3): image transitions is gray level image;

Step (4): image filtering;

Step (5): reading images address;

Step (6): calculate the gray scale difference value absolute average;

Step (7): whether judging the gray scale difference value absolute average greater than threshold value, is 255 if just set gray-scale value, and enters step (8); Be 0 if not just setting gray-scale value, enter step (8);

Step (8): judge whether image is handled, if just carry out edge extracting, the output image edge; Return to step (5) after just image address being added one.

The figure filtering of described step (4) is as follows: with Gaussian filter and median filter, image is carried out smothing filtering respectively, effectively suppress salt-pepper noise and Gaussian noise in image.

The concrete steps of the calculating gray scale difference value mean value of described step (6) are:

Absolute mean is asked in image pointwise after smothing filtering.Suppose g ₀Be the point of considering, select 3 * 3 neighborhoods, computing formula is as follows:

G _i＝|I(g ₀)-I(g _i)|

$G=\frac{\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{G}_i}{8}$

g1	g8	g7
			g2	g0	g6
g3	g4	g5

In formula, i is pixel, and span is 0～8; Gi represents the absolute difference of g0 point and gi point gray scale, I(gi) is the gray-scale value of pixel gi.G is at g ₀The response at place, this value representation g ₀The mean difference of the gray scale at some place and neighborhood territory pixel gray scale around it responds the character that G has embodied pixel.

The threshold value of described step (7) is chosen: according to the statistical property of image absolute mean matrix, carry out the threshold value of rim detection and choose

$\mathrm{avr}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\left|G(i,j)\right|}{M\×N}$

$\mathrm{std}=\sqrt{\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{(G(i,j)-\mathrm{avr})}^2}{M\×N}}$

T＝a×avr+b×std(a=1,b=0.5)

Wherein T represents selected threshold value, and M * N is picture size, and i, j are pixel, and span is respectively 1～M and 1～N; G (i, j) is the Grad that point (i, j) is located, and std and avr represent respectively standard deviation and the average of gradient matrix, and a, b are empirical constant, general a=1, b=0.5.When a, b get less value, can obtain more edge details; When a, b get larger value, can reduce the noise in edge image.

The edge extracting of described step (8):

Whole image is carried out single pass, and response is the marginal point of image greater than the point of threshold value, only by threshold values T just with Edge extraction out.

As shown in Figure 2, the original image before the present invention processes.

As shown in Figure 3, the image after the present invention processes.Can be found out by Fig. 2, Fig. 3, the method detected image border continuity is better, and pseudo-edge and isolated point less; Differentiate owing to not relating to, so do not increase noise in the process of figure image intensifying.

Although above-mentionedly by reference to the accompanying drawings the specific embodiment of the present invention is described; but be not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various modifications that creative work can make or distortion still in protection scope of the present invention.

Claims (5)

1. the method for detecting image edge that the intensity-based difference is average, is characterized in that, mainly comprises the steps:

Step (1): beginning, reading images;

Step (2): set gray threshold;

Step (3): image transitions is gray level image;

Step (4): image filtering;

Step (5): reading images address;

Step (6): calculate the gray scale difference value absolute average;

Step (7): whether judging the gray scale difference value absolute average greater than threshold value, is 255 if just set gray-scale value, and enters step (8); Be 0 if not just setting gray-scale value, enter step (8);

Step (8): judge whether image is handled, if just carry out edge extracting, the output image edge; Return to step (5) after just image address being added one.

2. the average method for detecting image edge of a kind of intensity-based difference as claimed in claim 1, it is characterized in that, the figure filtering of described step (4) is as follows: with Gaussian filter and median filter, image is carried out smothing filtering respectively, effectively suppress salt-pepper noise and Gaussian noise in image.

3. the average method for detecting image edge of a kind of intensity-based difference as claimed in claim 1, is characterized in that, the concrete steps of the calculating gray scale difference value mean value of described step (6) are:

Absolute mean is asked in image pointwise after smothing filtering, supposed g ₀Be the point of considering, select 3 * 3 neighborhoods, computing formula is as follows:

G _i＝|I(g ₀)-I(g _i)|

$G=\frac{\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{G}_i}{8}$

In formula, i is pixel, and span is 0～8; Gi represents the absolute difference of g0 point and gi point gray scale, I(gi) is the gray-scale value of pixel gi, and G is at g ₀The response at place, this value representation g ₀The mean difference of the gray scale at some place and neighborhood territory pixel gray scale around it responds the character that G has embodied pixel.

4. the average method for detecting image edge of a kind of intensity-based difference as claimed in claim 1, is characterized in that, the threshold value of described step (7) is chosen: according to the statistical property of image absolute mean matrix, carry out the threshold value of rim detection and choose

$\mathrm{avr}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\left|G(i,j)\right|}{M\×N}$

$\mathrm{std}=\sqrt{\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{(G(i,j)-\mathrm{avr})}^2}{M\×N}}$

T＝a×avr+b×std(a=1,b=0.5)

Wherein T represents selected threshold value, and M * N is picture size, and i, j are pixel, and the span of i and j is respectively 1～M and 1～N; G (i, j) is the Grad that point (i, j) is located, and std and avr represent respectively standard deviation and the average of gradient matrix, and a, b are empirical constant, when a, b get less value, obtains more edge details; When a, b get larger value, reduce the noise in edge image, preferred a=1, b=0.5.

5. the average method for detecting image edge of a kind of intensity-based difference as claimed in claim 1, it is characterized in that, the edge extracting of described step (8): whole image is carried out single pass, response is the marginal point of image greater than the point of threshold value, only by threshold values T just with Edge extraction out.

Images