CN105894474A - Non-linear image enhancement method, and edge detection method using the same - Google Patents

Abstract

The invention discloses a non-linear image enhancement method, and an edge detection method using the same. The non-linear image enhancement method includes the following steps: image conversion, image filtering and image equalization. The edge detection method also includes the steps: image denoising, gradient calculating, marking of non-edge pixels, and fine edge treatment. The non-linear image enhancement method, and the edge detection method using the same utilize the non-linear image filtering and equalization method to effectively eliminate image noises and enhance image effect, can overcome the limitation of a linear method, are higher in practicality, and can perform edge extraction after image enhancement, wherein the employed edge extraction algorithm has better detection capability and positioning capability and has minimum response so that the image contour can be extracted more clearly.

Description

A kind of nonlinear images Enhancement Method and edge detection method

Technical field

The present invention relates to a kind of image enchancing method and edge detection method, especially a kind of nonlinear images Enhancement Method and edge detection method, belong to technical field of image processing.

Background technology

In recent years, developing rapidly with universal along with multimedia technology and the Internet, image processing techniques ever more important, it is widely used in the fields such as office automation, industrial robot, Geoprocessing, earth resource monitoring, remote sensing, medical science, communication, interactive CAD, the most gradually comes in daily life.But owing to the conditions such as, environment noise uneven by light are affected, the effect of image does not complies with one's wishes, affect features is correct cognitive, therefore, strengthens image and the intensive process such as edge extracting is the most necessary.

The linear method of image enhaucament in traditional image and information processing the most all in occupation of basis, the status of core, and the physical process of reality is the most nonlinear, the method utilizing linear approximation cannot resolve its main character under many circumstances, thus to the analysis of the non-linear factors such as the morphological characteristic of image and geometry by system linearity feature and describe and have its limitation the most unavoidably.

First order differential operator such as Sobel operator, Roberts operator, Prewitt operator, Kirsch operator, when seeking the gradient at edge, need to calculate each pixel, and be typically necessary threshold value is previously set, and the edge obtained relatively " thick ", positions inaccurate.The edge detection operator of second-order differential such as Laplace operator and LoG operator, the most sensitive to noise, and due to zero cross point not with marginal point one_to_one corresponding, thus often can produce some false edges, thus limit its application.

Summary of the invention

For defect or the deficiency of above-mentioned prior art, the present invention proposes a kind of nonlinear images Enhancement Method and edge detection method.

For achieving the above object, the technical solution used in the present invention is as follows:

Technical scheme one:

A kind of nonlinear images Enhancement Method, comprises the following steps:

Step 1: image is changed: the coloured image of acquisition is converted into gray level image；

Step 2: image filtering: use median filter method that described gray level image is filtered:

G (x, y)=Med{f (x-k, y-l) }, (k, l ∈ W) (1)

Wherein, (x, y) is original image to f, and (x, is y) image after medium filtering to g, and W is two dimension pattern plate, and Med is for extracting median operation functional symbol；

Step 3: image equilibration: be made up of following sub-step:

Step 3-1: the frequency that before calculating equalization processing, image kth gray level occurs:

p_s(s_k)=n_k/n,1≤k≤N (2)

Wherein, n_kFor kth gray-level pixels number, n is sum of all pixels, and N is gray level sum；

Step 3-2: the gray value cumulative distribution function of gray level image before calculating equalization processing:

$t_i=\underset{k=0}{\overset{i}{\Σ}}{p}_s\left(s_k\right),0\≤i\≤N---\left(3\right)$

Step 3-3: adjust pixel value in gray level image, after making equalization processing, the gray value cumulative distribution function of gray level image meets:

${t^{\′}}_i=\underset{k=0}{\overset{i}{\Σ}}\frac{M}{N}\·i,0\≤i\≤N---\left(4\right)$

Wherein, M is the sum of all pixels of gray level image.

The template that two dimension pattern plate W is 5x5 size that described median filter method uses.

Technical scheme two:

A kind of use the edge detection method of nonlinear images Enhancement Method described in technical scheme one, also include that edge detecting step, described edge detecting step include following sub-step:

Step a: image denoising: gray level image and 2-d gaussian filters template are carried out convolution algorithm, eliminates noise；

Step b: gradient calculation: utilize derivative operator to calculate gradient | G | and orientation angle θ thereof of each edge pixel in gray level image:

$\left|G\right|=\sqrt{G_x^2+G_y^2}---\left(5\right)$

$\mathrm{\θ}=arctan\left(\frac{G_y}{G_x}\right)---\left(6\right)$

Wherein, G_xAnd G_yIt is respectively pixel at x direction and the derivative in y direction；

Step c: labelling non-edge pixels: according to gradient direction, centered by each pixel, determine the adjacent pixels of its both sides；If the gray value of current pixel is not maximum compared with the gray value of its two adjacent pixels, in the most described edge graph, current pixel is labeled as non-edge pixels；

Step d: edge process of refinement: each edge pixel in detection gray level image one by one；If the gray value of current pixel is less than Low threshold, then respective pixel in edge graph is labeled as non-edge pixels；If the gray value of current pixel is less than or equal to high threshold and more than or equal to Low threshold, and there is not the adjacent pixels more than high threshold in it, then respective pixel in edge graph is labeled as non-edge pixels.

High threshold and Low threshold in described step d are determined by accumulative histogram.

The beneficial effects of the present invention is:

The present invention uses nonlinear image filtering and equalization method, effectively eliminate picture noise and strengthen image effect, overcome the limitation of linear method, there is higher practicality, and after image enhaucament, carry out edge extracting, the Boundary extracting algorithm used has preferably detection property, polarization and minimum response, makes image outline more clearly be extracted.

Accompanying drawing explanation

Fig. 1 is the flow chart of the embodiment of the present invention 1；

Fig. 2 is the gray level image in the embodiment of the present invention 1；

Fig. 3 is the gray level image in the embodiment of the present invention 1 through medium filtering；

Fig. 4 is the rectangular histogram of gray level image before equalization processing in the embodiment of the present invention 1；

Fig. 5 is the rectangular histogram of gray level image in the embodiment of the present invention 1 through equalization processing；

Fig. 6 is the gray level image in the embodiment of the present invention 1 through equalization processing；

Fig. 7 is the flow chart of the embodiment of the present invention 2；

Fig. 8 is the edge graph of the embodiment of the present invention 2.

Detailed description of the invention

Embodiment 1:

As it is shown in figure 1, a kind of nonlinear images Enhancement Method, comprise the following steps:

Step 1: image is changed: the coloured image of acquisition is converted into gray level image；

Step 2: image filtering: use median filter method that described gray level image is filtered；

Step 3: image equilibration: be made up of following sub-step:

Step 3-1: the frequency that before calculating equalization processing, image kth gray level occurs:

p_s(s_k)=n_k/n,1≤k≤N (1)

Wherein, n_kFor kth gray-level pixels number, n is sum of all pixels, and N is gray level sum；

Step 3-2: the gray value cumulative distribution function of gray level image before calculating equalization processing:

$t_i=\underset{k=0}{\overset{i}{\Σ}}{p}_s\left(s_k\right),0\≤i\≤N---\left(2\right)$

Step 3-3: adjust pixel value in gray level image, after making equalization processing, the gray value cumulative distribution function of gray level image meets:

${t^{\′}}_i=\underset{k=0}{\overset{i}{\Σ}}\frac{M}{N}\·i,0\≤i\≤N---\left(3\right)$

Wherein, M is the sum of all pixels of gray level image.

When processing coloured image, respectively tri-kinds of components of RGB are processed.But actually RGB can not reflect the morphological characteristic of image, simply carrying out the allotment of color from the principle of optics, therefore, image carries out gray processing to reduce amount of calculation, the gray level image after conversion is as shown in Figure 2.

If input picture contains noise, the effect of image and the result of rim detection will be affected, therefore, need image is filtered, median filter method is a kind of method that image is carried out Nonlinear Processing, and it substitutes the intermediate value of each point value in one neighborhood of this point of the value of any in digital picture, allows the pixel value of surrounding close to actual value, thus eliminating isolated noise spot, the image that medium filtering processed is as shown in Figure 3.

The concrete methods of realizing of medium filtering is the two-dimentional sleiding form with ad hoc structure, is ranked up by size by pixel value in template, extracts sequence numerical value placed in the middle as filtering output value, and its mathematic(al) representation is:

G (x, y)=Med{f (x-k, y-l) }, (k, l ∈ W) (4)

Wherein, (x, y) is original image to f, and (x, is y) image after medium filtering to g, and W is two dimension pattern plate, and Med is for extracting median operation functional symbol.The rectangle template that described median filter method uses window to be 5x5 size.

The frequency that kth gray level in gray level image before equalization processing occurs is fitted to curve, and being equalized processes the rectangular histogram of front gray level image, as shown in Figure 4；Become by the rectangular histogram of the gray level image of equalization processing and be uniformly distributed, as shown in Figure 5；In gray level image after equalization processing, the dynamic range of image intensity value increases, thus reaches to strengthen image overall contrast, make image become effect clearly, as shown in Figure 6.

Embodiment 2:

As it is shown in fig. 7, the edge detection method of a kind of nonlinear images Enhancement Method used described in claim 1, it is characterised in that: also include edge detecting step: described edge detecting step includes following sub-step:

Step a: image denoising: gray level image and 2-d gaussian filters template are carried out convolution algorithm, eliminates noise；

Step b: gradient calculation: utilize derivative operator to calculate gradient | G | and orientation angle θ thereof of each edge pixel in gray level image:

$\left|G\right|=\sqrt{G_x^2+G_y^2}---\left(5\right)$

$\mathrm{\θ}=arctan\left(\frac{G_y}{G_x}\right)---\left(6\right)$

Wherein, G_xAnd G_yIt is respectively pixel at x direction and the derivative in y direction；

Step c: labelling non-edge pixels: according to gradient direction, centered by each pixel, determine the adjacent pixels of its both sides；If the gray value of current pixel is not maximum compared with the gray value of its two adjacent pixels, in the most described edge graph, current pixel is labeled as non-edge pixels；

Step d: edge process of refinement: each edge pixel in detection gray level image one by one；If the gray value of current pixel is less than Low threshold, then respective pixel in edge graph is labeled as non-edge pixels；If the gray value of current pixel is less than or equal to high threshold and more than or equal to Low threshold, and there is not the adjacent pixels more than high threshold in it, then respective pixel in edge graph is labeled as non-edge pixels.

High threshold and Low threshold in described step d are determined by accumulative histogram.

The edge graph extracted is as shown in Figure 8.

It should be noted that without departing under present inventive concept premise, any minor variations being the present invention and modification belong to protection scope of the present invention.

Claims (4)

1. a nonlinear images Enhancement Method, it is characterised in that: comprise the following steps:

Step 1: image is changed: the coloured image of acquisition is converted into gray level image；

Step 2: image filtering: use median filter method that described gray level image is filtered:

G (x, y)=Med{f (x-k, y-l) }, (k, l ∈ W) (1)

Wherein, (x, y) is original image to f, and (x, is y) image after medium filtering to g, and W is two dimension pattern plate, Med For extracting median operation functional symbol；

Step 3: image equilibration: be made up of following sub-step:

Step 3-1: the frequency that before calculating equalization processing, image kth gray level occurs:

p_s(s_k)=n_k/ n, 1≤k≤N (2)

Wherein, n_kFor kth gray-level pixels number, n is sum of all pixels, and N is gray level sum；

Step 3-2: the gray value cumulative distribution function of gray level image before calculating equalization processing:

$t_i=\underset{k=0}{\overset{i}{\Σ}}{p}_s\left(s_k\right),0\≤i\≤N---\left(3\right)$

Step 3-3: adjust pixel value in gray level image, after making equalization processing, the gray value of gray level image tires out Long-pending distribution function meets:

${t^{\′}}_i=\underset{k=0}{\overset{i}{\Σ}}\frac{M}{N}\·i,0\≤i\≤N---\left(4\right)$

Wherein, M is the sum of all pixels of gray level image.

Nonlinear images Enhancement Method the most according to claim 1, it is characterised in that: described intermediate value The template that two dimension pattern plate W is 5 × 5 sizes that filtering method uses.

3. use an edge detection method for nonlinear images Enhancement Method described in claim 1, its It is characterised by: also include edge detecting step: described edge detecting step includes following sub-step:

Step a: image denoising: gray level image and 2-d gaussian filters template are carried out convolution algorithm, and elimination is made an uproar Sound；

Step b: gradient calculation: utilize derivative operator calculate each edge pixel in gray level image gradient | G | and Its orientation angle θ:

$\left|G\right|=\sqrt{G_x^2+G_y^2}---\left(5\right)$

$\mathrm{\θ}=arctan\left(\frac{G_y}{G_x}\right)---\left(6\right)$

Wherein, G_xAnd G_yIt is respectively pixel at x direction and the derivative in y direction；

Step c: labelling non-edge pixels: according to gradient direction, centered by each pixel, determine its both sides Adjacent pixels；If the gray value of current pixel is not maximum compared with the gray value of its two adjacent pixels, In the most described edge graph, current pixel is labeled as non-edge pixels；

Step d: edge process of refinement: each edge pixel in detection gray level image one by one；If current pixel Gray value is less than Low threshold, then respective pixel in edge graph is labeled as non-edge pixels；If current pixel Gray value is less than or equal to high threshold and more than or equal to Low threshold, and it does not exist the adjacent pixels more than high threshold, Then respective pixel in edge graph is labeled as non-edge pixels.

Edge detection method the most according to claim 3, it is characterised in that: the height in described step d Threshold value and Low threshold are determined by accumulative histogram.