WO2014173145A1 - 一种图像增强方法及设备 - Google Patents

一种图像增强方法及设备 Download PDF

Info

Publication number
WO2014173145A1
WO2014173145A1 PCT/CN2013/089404 CN2013089404W WO2014173145A1 WO 2014173145 A1 WO2014173145 A1 WO 2014173145A1 CN 2013089404 W CN2013089404 W CN 2013089404W WO 2014173145 A1 WO2014173145 A1 WO 2014173145A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
pixel
neighborhood
pixel point
suppression
Prior art date
Application number
PCT/CN2013/089404
Other languages
English (en)
French (fr)
Inventor
杜馨瑜
王栋
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP13883013.8A priority Critical patent/EP2983130A4/en
Publication of WO2014173145A1 publication Critical patent/WO2014173145A1/zh
Priority to US14/920,669 priority patent/US9704225B2/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/73Deblurring; Sharpening
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/20Image enhancement or restoration using local operators
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/40Image enhancement or restoration using histogram techniques
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/90Dynamic range modification of images or parts thereof
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/90Determination of colour characteristics
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20172Image enhancement details
    • G06T2207/20192Edge enhancement; Edge preservation

Definitions

  • the present invention relates to the field of image processing, and in particular, to an image enhancement method and apparatus. Background technique
  • Image enhancement technology is one of the key technologies in the field of image processing. It is used to improve and enhance the original image quality, and even reveal the hidden information in the original image, making it more suitable for the observation of human visual systems or the processing of other functional modules. . And image enhancement technology has important applications in the fields of remote sensing, dynamic scene analysis, automatic navigation, and medical image analysis.
  • image enhancement by gamma correction which is based on the power conversion of the input image pixel grayscale, as officially shown:
  • c is the normal number
  • r and 5 are the input image and output image pixel grayscale, respectively.
  • the embodiment of the invention provides an image enhancement method and device, which can achieve enhanced detail of the enhanced image.
  • an image enhancement method provided by an embodiment of the present invention includes:
  • the neighborhood pixel point of the target pixel point is specified, and the suppression value of the neighboring pixel point to the target pixel point refers to a value represented by the suppression effect of the neighborhood pixel point on the target pixel point;
  • the value of 0 is taken as the pixel value of the target pixel after the image is enhanced.
  • an image enhancement device including:
  • the first calculating module is configured to calculate a center response value of the target pixel point based on an original pixel value of the target pixel point, where the center response value is greater than the original pixel value;
  • the second calculating module is configured to calculate a first total suppression value of the target pixel point, where the first total suppression value is a suppression value of all the neighboring pixel points of the target pixel point to the target pixel point a neighboring pixel point is a neighboring pixel point previously designated as the target pixel point, and the suppression value of the neighboring pixel point to the target pixel point refers to the neighboring pixel point to the target pixel point
  • the image enhancement module is configured to: when the center response value is greater than the first total suppression value, subtract the first total suppression value from the central response value to obtain a difference, and use the difference as a pixel value of the target pixel after image enhancement; and a pixel value of the target pixel after the image is enhanced as the center response value is less than the first total suppression value.
  • the first total suppression value is a sum of suppression values of all the neighboring pixel points of the target pixel to the target pixel, and the neighboring pixel is a neighboring pixel previously designated as the target pixel
  • the suppression value of the neighboring pixel point to the target pixel point refers to a value represented by the suppression effect of the neighborhood pixel point on the target pixel point; when the center response value is greater than the first total suppression value And subtracting the first total suppression value from the central response value to obtain a difference, and using the difference as a pixel value of the target pixel after image enhancement; when the central response value is smaller than the first When the total suppression value is
  • the value of 0 is the pixel value of the target pixel after the image is enhanced. This increases the image due to the target pixel Strongly determined by the original pixel value of the target pixel and the original pixel value of the other neighboring pixel points, so that only a simple power conversion is performed on the pixel value of the pixel point compared with the prior art, and the image is enhanced in the embodiment of the present invention. The image is more detailed. DRAWINGS
  • FIG. 1 is a schematic flowchart of an image enhancement method according to an embodiment of the present invention.
  • FIG. 2 is a schematic flow chart of another image enhancement method according to an embodiment of the present invention.
  • FIG. 3 is a schematic flowchart of another image enhancement method according to an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of an optional calculation template according to an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of comparison of effects of various image enhancements according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram showing another effect of comparing various image enhancements according to an embodiment of the present invention
  • FIG. 7 is a schematic diagram showing another effect of comparing various image enhancements according to an embodiment of the present invention
  • FIG. 8 is a schematic diagram of an embodiment of the present invention. Schematic diagram of an image enhancement device
  • FIG. 9 is a schematic structural diagram of another image enhancement device according to an embodiment of the present invention.
  • FIG. 10 is a schematic structural diagram of another image enhancement device according to an embodiment of the present invention
  • FIG. 11 is a schematic structural diagram of another image enhancement device according to an embodiment of the present invention. detailed description
  • FIG. 1 is a schematic flowchart of an image enhancement method according to an embodiment of the present invention. As shown in FIG. 1, the method includes:
  • the first total suppression value is a sum of suppression values of all neighboring pixel points of the target pixel to the target pixel, the neighborhood
  • the pixel is a neighboring pixel that is pre-designated as the target pixel, and the suppression value of the neighboring pixel to the target pixel is represented by the suppression of the target pixel by the neighboring pixel.
  • the target pixel may be one or more pixels in an image that needs to be image enhanced.
  • image enhancement shown in the above step may be performed on other pixels in the image, that is, the image enhancement shown in the above step may be performed on the entire image.
  • the method can also be applied to any device that supports image processing, such as a computer, a mobile phone, a tablet, and the like.
  • the first total suppression value is a sum of suppression values of all the neighboring pixel points of the target pixel to the target pixel, and the neighboring pixel is a neighboring pixel previously designated as the target pixel
  • the suppression value of the neighboring pixel point to the target pixel point refers to a value represented by the suppression effect of the neighborhood pixel point on the target pixel point; when the center response value is greater than the first total suppression value And subtracting the first total suppression value from the central response value to obtain a difference, and using the difference as a pixel value of the target pixel after image enhancement; when the central response value is smaller than the first
  • the value of 0 is taken as the pixel value of the target pixel after the image is enhanced.
  • FIG. 2 is a schematic flowchart diagram of another image enhancement method according to an embodiment of the present invention, as shown in FIG. 2 As shown, including:
  • a second total suppression value of each of the neighboring pixel points where a second total suppression value of the neighboring pixel points is a sum of suppression values of the other neighboring pixel points to the neighboring pixel points,
  • the other neighboring pixel points refer to all neighboring pixel points of the all neighboring pixel points except the neighboring pixel points.
  • calculating a second suppression value of one of the neighboring pixel points is calculating a sum of the suppression values of the other seven neighboring pixel points to the neighboring pixel points
  • the sum of the other seven suppression values for the pixel of the neighborhood may also be understood as the value indicated by the total suppression of the neighboring pixel points by the other seven neighborhood pixels.
  • a suppression value of the neighboring pixel point to the target pixel point is The difference between the suppression points of the neighboring pixel points is the difference between the original pixel value of the neighboring pixel point minus the second total suppression value of the neighboring pixel point.
  • a value, or a suppression difference value of the neighboring pixel point is 0, where the T is calculated based on an original pixel value of the neighboring pixel point, and a distance value between the neighboring pixel point and the target pixel point The value that comes out.
  • step 202 calculates a second total suppression value of each neighborhood pixel, so that step 203 can calculate a suppression difference of each neighborhood pixel by using the second total suppression value, and then A T calculated by the original pixel value of the neighboring pixel and the distance between the neighboring pixel and the target pixel, and the suppression value of each of the neighboring pixel points to the target pixel is obtained.
  • the above T can be a function, or a real value.
  • step 203 may include:
  • a potential function of the neighboring pixel point is a norm value of a difference between an original pixel value of the neighboring pixel point and an original pixel value of the target pixel point ;
  • the rate of change of the potential function being a ratio of a power function of the potential function to a power function of all potential functions, the power function of the potential function
  • the exponent is the potential function
  • the base of the power function of the potential function is a preset base
  • Calculating a kernel value of each of the neighboring pixel points, and the kernel value of the neighboring pixel point is based on the neighborhood a rate of change of a potential function of the pixel, and a value calculated from a distance value between the neighboring pixel point and the target pixel point;
  • the T difference of the suppression point difference of the domain pixel is the difference between the original pixel value of the neighboring pixel point minus the second total suppression value of the neighboring pixel point, Or the difference between the suppression of the neighboring pixel points is 0, and the T is a value calculated based on the original pixel value of the neighboring pixel point and the kernel value of the neighboring pixel point.
  • the above potential function can be as follows:
  • D ( x , a potential function representing a neighboring pixel point (X, y) may be equal to 1 or equal to 2, etc., that is, the above may be a norm or a 2 norm, etc.
  • the above represents a neighboring pixel point.
  • the pixel value of (X, y), and the above ( ⁇ .) represents the pixel value of the target pixel.
  • the rate of change of D ( x ) can be as follows:
  • x above can represent the rate of change of D ( x , ) of the neighborhood pixel (X, y), and the above ex p(Z)(x, ) can represent the natural logarithm e
  • w represents a set of all neighboring pixel points, that is, the sum of the ex P( D ( x , )) of all the neighboring pixel points is obtained in the above formula.
  • the above core values can be as follows:
  • the above represents a kernel value of a neighboring pixel point (X, y), wherein the above is 1/H times a distance value between the neighboring pixel point (X, y) and the target pixel point, wherein H is a preset
  • H is a preset
  • the fixed value, for example, H can be set to 3.
  • the T in the embodiment may be a value calculated based on an original pixel value of the neighboring pixel point and a kernel value of the neighboring pixel point, and the calculation of the kernel value is based on different neighboring pixel points.
  • the interaction between the target pixel and the neighboring pixel is also combined, and the local contrast information composed of the target pixel and the adjacent pixel is used to perform the above T Full scale adjustment.
  • the image enhancement implemented by the embodiment in the embodiment of the present invention is defined as a Variational Scalar non-classical Receptive Field (VSNRF).
  • VSNRF Variational Scalar non-classical Receptive Field
  • the image enhancement achieved is defined as basic.
  • step 205 can be implemented by the following formula:
  • T represents a pixel value of the target pixel after image enhancement, and represents a center response value calculated in step 201, and represents a first total suppression value calculated in step 204.
  • the max operation described above makes the total response non-negative and implies a non-linear effect, and is a Kroneck function.
  • the method may further include:
  • the template includes a central area, and each of the central areas is provided with a neighborhood area;
  • the target pixel point refers to a pixel point covered by the central area, and the target pixel
  • the neighboring pixel of the point is the pixel point covered by the neighborhood of the central area.
  • the calculation template may be a template of 3 ⁇ 3, 5 ⁇ 5, 7 ⁇ 7, etc., wherein the number of central regions included in different templates may be different, for example, the template of 3 ⁇ 3 includes one In the central area, a 5 x 5 template can contain one or more central areas.
  • the size of the central area may be the same as the size of one pixel in the image, that is, the target pixel is one; the size of the central area may be the same as the size of several pixels in the image, that is, the target. There are multiple pixels.
  • the neighborhood area of the central area has an uncovered pixel point neighborhood area that does not cover a pixel point in the image where the target pixel point is located, that is, the target pixel point is an edge image of the image.
  • the neighboring pixel points of the target pixel covered by the central area include:
  • the pixel points in the image covered by the neighborhood area of the central area is set to 0.
  • the template of 3 3 can be as shown in Fig. 4, in which a black circle represents a central area, and the other eight white circles represent a neighborhood area.
  • a coordinate system is set in the template, and the origin of the coordinate system is the center area, that is, (x, _y.), and the dotted line in FIG. 4 shows the coordinate values of other neighborhood pixel points ( x , the suppression value of the neighborhood pixel point).
  • the sum of all the dashed lines is the second total suppression value of the neighborhood pixel ( x , and the solid line in Fig. 4 represents the suppression value of the neighborhood pixel ( x , the target pixel (x, _y.)).
  • step 201 may include:
  • step 201 may be calculated by the center of the target pixel value in response to the following formula: (. X .J) where the center pixel represents the target response value,
  • (w.) indicates the original pixel value of the target pixel point ( ⁇ , y0).
  • step 202 can include:
  • the second total suppression value of each of the neighborhood pixel points is calculated by the following formula:
  • represents the second total suppression value of the neighborhood pixel ( x , ⁇ ( x , represents the removal of the neighborhood pixel ( x , a set of all neighborhood points;
  • ⁇ , ⁇ + ") 4 (dur3) H represents the suppression value of the neighborhood pixel (x + ⁇ , _y + ") to the neighborhood pixel y)
  • step 203 may include:
  • the suppression value of each of the neighboring pixel points to the target pixel point is calculated by the following formula:
  • denotes the neighborhood pixel point ( x , the suppression value of the target pixel point ( ⁇ , ⁇ ), 4 is the sensitivity of the neighborhood pixel point to the target pixel point, and is the neighborhood pixel point
  • is the Gaussian function parameter
  • the maximum distance between the template center point and the neighboring pixel point shown in Figure 4 is 7 ⁇ .
  • the max operation makes the suppression effect non-negative, implying a nonlinear effect. For the Kroneck function.
  • ⁇ 2 in the above formula may be as described above.
  • BNRF the above equation may be equal to ⁇ 2 is equal to the target pixel ( ⁇ . ⁇ ) and the neighborhood pixel point (x, H times the distance values.
  • step 204 can include:
  • the sum of the suppression values of all the neighboring pixel points to the target pixel point is calculated as the first total suppression value of the target pixel point by the following formula:
  • N the set of neighborhood points, and there are 8 neighborhood points in this template.
  • the embodiment may further compare the image enhanced image implemented by the embodiment with the image enhanced image implemented by other methods.
  • the image is an image with uneven brightness distribution.
  • Figure 5 A is the original image
  • B is the image enhanced by the histogram equalization (HE) technique
  • C is the image enhancement of the original image by Retinex (a well-known image algorithm) algorithm.
  • the image, D is an image obtained by image enhancement of the original image by BNRF
  • E is an image obtained by image enhancement of the original image by VSNRF.
  • VSNRF shows more details, especially the high-brightness area of the moon.
  • the E map is not lost. It can be seen that the image enhancement effect of VSNRF is the best among the above several techniques.
  • the method may further evaluate the image after the image is enhanced, and the image-enhanced image may be evaluated by using at least one of the following indicators:
  • EC Evaluation of Contrast
  • EME Measure of Enhancement
  • IP Intensity Preservation
  • EC(g) log(log( ⁇ lG r (", m ) 2 + G c (", m) 2 )) is the above-mentioned EC index, and is a calculation template set in this embodiment in the horizontal and vertical directions (for example)
  • the IP indicator is used to avoid a significant difference in the overall perception between the enhanced image and the original image, especially to avoid excessive contrast in the sequence image.
  • IP to represent the difference between the mean of the original image and the average of the enhanced image. The closer to 0, the better the gray level of the corresponding image enhancement algorithm is maintained.
  • the original image is an image in a database, as shown in the leftmost column of the image in Figure 6. From left to right, the results of the HE technique, the results of the Retinex technique, the results of the BNRF technique and the results of the VSNRF technique are shown. Table 1 shows the results of the above methods measured by the three indicators of EC, EME and IP. It can be seen that the image enhancement achieved by VSNRF is superior to the image enhancement achieved by other techniques.
  • the A, B, and C diagrams in Figure 7 show the Mach effect.
  • the A diagram in Fig. 7 is a stepped original diagram
  • the B diagram in Fig. 7 is a result diagram after being processed by the embodiment of the present invention. It can be seen that Figure 7
  • the B picture in the middle shows a darker line in the dark area of the stepped edge. Similarly, a brighter line appears in the bright area, which is the Mach phenomenon in the optical illusion.
  • the C diagram in Fig. 7 is a side view of the result graph, and it can be clearly seen that the Mach effect is a spike of a protrusion.
  • the D and E diagrams in Figure 7 show the checkerboard effect.
  • the D picture in Fig. 7 is an original picture with a checkerboard pattern.
  • the human eye can see dark circles at the intersection of white, which is the checkerboard effect in the optical illusion.
  • the E diagram in Fig. 7 is a cross-sectional view corresponding to the two regions after the side view processed by the embodiment of the present invention, and it can be seen that the curve of the sudden depression can explain the illusion of the circular spot appearing by the checkerboard effect.
  • FIG. 8 is a schematic structural diagram of an image enhancement device according to an embodiment of the present invention. As shown in FIG. 8, the method includes: a first calculation module 31, a second calculation module 32, and an image enhancement module 33, wherein:
  • the first calculating module 31 is configured to calculate a center response value of the target pixel point based on an original pixel value of the target pixel point, where the center response value is greater than the original pixel value;
  • a second calculating module 32 configured to calculate a first total suppression value of the target pixel point, where the first total suppression value is a suppression value of all the neighboring pixel points of the target pixel point to the target pixel point a neighboring pixel point is a neighboring pixel point previously designated as the target pixel point, and the suppression value of the neighboring pixel point to the target pixel point refers to the neighboring pixel point to the target pixel point
  • the image enhancement module 33 is configured to: when the center response value is greater than the first total suppression value, subtract the first total suppression value from the central response value to obtain a difference, and use the difference as an image a pixel value of the target pixel after the enhancement; and a pixel value of the target pixel after the image is enhanced by the value of 0 when the center response value is less than the first total suppression value.
  • the target pixel point may be one or more pixels in the image that needs to be image enhanced.
  • image enhancement is performed on the target pixel, and in this embodiment, Performing the image enhancement shown in the above steps on the other pixels in the above image, that is, the image enhancement shown in the above steps can be performed on the entire image.
  • the device may be any device that supports image processing, such as a computer, a mobile phone, a tablet, and the like.
  • the first total suppression value is a sum of suppression values of all the neighboring pixel points of the target pixel to the target pixel, and the neighboring pixel is a neighboring pixel previously designated as the target pixel
  • the suppression value of the neighboring pixel point to the target pixel point refers to a value represented by the suppression effect of the neighborhood pixel point on the target pixel point; when the center response value is greater than the first total suppression value And subtracting the first total suppression value from the central response value to obtain a difference, and using the difference as a pixel value of the target pixel after image enhancement; when the central response value is smaller than the first
  • the value of 0 is taken as the pixel value of the target pixel after the image is enhanced.
  • FIG. 9 is a schematic structural diagram of another image enhancement device according to an embodiment of the present invention.
  • the method includes: a first calculation module 41, a second calculation module 42, and an image enhancement module 43.
  • the second calculation module includes a calculation submodule 421, a second calculation submodule 422, and a third calculation submodule 423, wherein:
  • the first calculating module 41 is configured to calculate a center response value of the target pixel point based on an original pixel value of the target pixel point, where the center response value is greater than the original pixel value.
  • a first sub-calculation sub-module 421, configured to separately calculate a second total suppression value of each of the neighboring pixel points, where the second total suppression value of the neighboring pixel points refers to other neighborhood pixel points to the neighborhood The sum of the suppression values of the pixel points, wherein the other neighborhood pixel points refer to all of the neighborhood pixel points except the neighborhood pixel point.
  • the neighboring pixels of the target pixel point comprise eight
  • calculating a second suppression value of one of the neighboring pixel points that is, calculating a total suppression value of the other seven neighboring pixel points to the neighboring pixel point.
  • the sum of the other seven suppression values for the pixel of the neighborhood may be understood as a value represented by the total suppression of the neighboring pixel points by the other seven neighborhood pixels.
  • a second sub-calculation sub-module 422 configured to calculate, according to a second total suppression value of each of the neighboring pixel points, a suppression value of the target pixel point to the target pixel point; the neighboring pixel point pair
  • the suppression value of the target pixel is T times the suppression difference of the neighboring pixel
  • the suppression difference of the neighboring pixel is the original pixel value of the neighboring pixel minus the neighboring pixel a difference obtained by the two total suppression values, or a suppression difference value of the neighboring pixel points is 0,
  • the T is an original pixel value based on the neighboring pixel point in advance, and the neighboring pixel point and the target The value calculated from the distance between the pixels.
  • the first sub-calculation sub-module 421 calculates a second total suppression value of each neighborhood pixel, so that the second sub-calculation sub-module 422 can calculate each neighborhood pixel by using the second total suppression value.
  • the suppression difference value is further calculated according to the original pixel value based on the neighboring pixel point and the distance T calculated by the distance between the neighboring pixel point and the target pixel point, to obtain each neighborhood pixel point pair.
  • the suppression value of the target pixel can be a function, or a real value.
  • the third sub-calculation sub-module 423 is configured to use a sum of the suppression values of all the neighboring pixel points to the target pixel point as the first total suppression value of the target pixel point.
  • the image enhancement module 43 is configured to: when the center response value is greater than the first total suppression value, subtract the first total suppression value from the central response value to obtain a difference, and use the difference as an image a pixel value of the target pixel after the enhancement; and a pixel value of the target pixel after the image is enhanced by the value of 0 when the center response value is less than the first total suppression value.
  • the image enhancement module 43 can also be used to achieve image enhancement of the target pixel by the following formula:
  • T represents a pixel value of the target pixel after image enhancement, and represents a center response value calculated in step 201, and represents a first total suppression value calculated in step 204.
  • the max operation described above makes the total response non-negative and implies a non-linear effect, and is a Kroneck function.
  • the second sub-computing sub-module 422 may further include: a first calculating unit (not shown in the drawing), configured to calculate a potential function of each of the neighboring pixel points, The potential function of the neighboring pixel is the norm of the difference between the original pixel value of the neighboring pixel point minus the original pixel value of the target pixel point; a second calculating unit (not shown in the drawing), configured to calculate a rate of change of a potential function of each of the neighboring pixel points, wherein the rate of change of the potential function refers to a power function of the potential function and all potential functions a ratio of a power function, an exponent of the power function of the potential function is the potential function, and a base of the power function of the potential function is a preset base;
  • a third calculating unit (not shown in the figure), configured to calculate a kernel value of each of the neighboring pixel points, where a kernel value of the neighboring pixel point is based on a rate of change of a potential function of the neighboring pixel point And a value calculated from a distance value between the neighboring pixel point and the target pixel point;
  • a fourth calculating unit (not shown in the figure), configured to calculate, according to a second total suppression value of each of the neighboring pixel points, a suppression value of the neighboring pixel point to the target pixel point; the neighboring pixel
  • the suppression value of the target pixel point is T times the suppression difference value of the neighborhood pixel point
  • the suppression difference value of the neighborhood pixel point is the original pixel value of the neighborhood pixel point minus the neighborhood a difference obtained by the second total suppression value of the pixel, or a suppression difference of the neighboring pixel point is 0,
  • T is an original pixel value based on the neighboring pixel point in advance
  • the neighboring pixel point The value calculated by the Gaussian function kernel.
  • the above potential function can be as follows:
  • D ( x , the potential function of the neighboring pixel point (X, y ) may be equal to 1 or equal to 2, that is, the above may be 1 norm or 2 norm, etc.
  • the above represents the neighboring pixel.
  • the pixel value of the point (X, y), and the above ( ⁇ .) represents the pixel value of the target pixel.
  • the rate of change of D ( x ) can be as follows:
  • x above can represent the rate of change of D ( x , ) of the neighborhood pixel (X , y ), and the above ex p(Z)(x, ) can represent the natural logarithm e
  • w represents a set of all neighboring pixel points, that is, the sum of the ex P( D ( x , )) of all the neighboring pixel points is obtained in the above formula.
  • the above core value can be calculated by the following formula
  • the above represents a kernel value of a neighboring pixel point (X, y), wherein the above is a 1/H times a distance value between the neighboring pixel point (X, y) and the target pixel point, where H is a preset Fixed value, For example, H can be set to 3.
  • the T in the embodiment may be a value calculated based on an original pixel value of the neighboring pixel point and a kernel value of the neighboring pixel point, and the calculation of the kernel value is based on different neighboring pixel points.
  • the interaction between the target pixel and the neighboring pixel is also combined, and the local contrast information composed of the target pixel and the adjacent pixel is used to make the above-mentioned ⁇ more sufficient scale. Adjustment.
  • the image enhancement achieved by this embodiment in the embodiment of the present invention is defined as VSNRF.
  • the image enhancement achieved is defined as basic. BNRF.
  • the device may further include: a setting unit 44, configured to set an image enhanced calculation template, where the template includes a central area, and for each of the central areas A neighborhood area is provided; the target pixel point refers to a pixel point covered by the central area, and a neighboring pixel point of the target pixel point is a pixel point covered by a neighborhood area of the central area.
  • a setting unit 44 configured to set an image enhanced calculation template, where the template includes a central area, and for each of the central areas A neighborhood area is provided; the target pixel point refers to a pixel point covered by the central area, and a neighboring pixel point of the target pixel point is a pixel point covered by a neighborhood area of the central area.
  • the calculation template may be a template of 3 ⁇ 3, 5 ⁇ 5, 7 ⁇ 7, etc., wherein the number of central regions included in different templates may be different, for example, the template of 3 ⁇ 3 includes one In the central area, a 5 x 5 template can contain one or more central areas.
  • the size of the central area may be the same as the size of one pixel in the image, that is, the target pixel is one; the size of the central area may be the same as the size of several pixels in the image, that is, the target. There are multiple pixels.
  • a neighborhood of the target pixel point covered by the central area Pixels include:
  • the following is an example of a 3 x 3 template as shown in FIG. 4 , in which a coordinate system is set, and the origin of the coordinate system is a central region, that is, (x., _y.),
  • the dotted line in 4 shows the coordinate values of other neighborhood pixel points ( x , the suppression value of the neighboring pixel points, and the sum of all the dotted lines is the neighborhood image
  • the second total suppression value of the prime point ( x , the solid line in Fig. 4 represents the suppression value of the neighborhood pixel point ( x , the target pixel point (x, _y.)).
  • the first calculation module 41 can also calculate the center response value of the target pixel by the following formula:
  • J o represents the original pixel value of the target pixel point ( ⁇ ⁇ , y0 ).
  • the first calculation sub-module 421 can also be used to calculate a second total suppression value for each of the neighborhood pixel points by the following formula:
  • the second calculation sub-module 422 calculates the suppression value of each of the neighborhood pixels to the target pixel by the following formula:
  • denotes the neighborhood pixel point ( x , the suppression value of the target pixel point ( ⁇ , ⁇ ), 4 is the sensitivity of the neighborhood pixel point to the target pixel point, and is the neighborhood pixel The pixel value of the point.
  • is high
  • the max operation makes the inhibition a non-negative value, implying a nonlinear effect.
  • ⁇ ) is the Kroneck function.
  • ⁇ 2 in the above formula may be as described above.
  • BNRF ⁇ 2 in the above equation may be equal to the target pixel (X .J.) And the neighborhood pixel point (x, from the value 1 / H times.
  • the third calculation sub-module 423 can also be used to calculate the sum of the suppression values of all the neighborhood pixel points to the target pixel point as the first total suppression of the target pixel point by the following formula Value:
  • N the set of neighborhood points, and there are 8 neighborhood points in this template.
  • the device may perform image enhancement processing only on gray or black and white images, and may perform image enhancement processing on the color image, but the color image may be sequentially performed on the RGB three-way by the above method. Image enhancement processing.
  • FIG. 9 is a schematic structural diagram of another image enhancement device according to an embodiment of the present invention.
  • a memory 51 and a processor 52 connected to the memory 51, the memory 51 is configured to store a set of program codes, a processor. 52 is used to call the program stored in the memory 51 to perform the following operations:
  • the first total suppression value is a sum of suppression values of all neighborhood pixel points of the target pixel point to the target pixel point, the neighborhood pixel point a neighborhood pixel that is pre-designated as the target pixel, and the suppression value of the neighborhood pixel to the target pixel refers to a value represented by the suppression of the target pixel by the neighborhood pixel;
  • the first total suppression value is obtained as a difference, and the difference is used as a pixel value of the target pixel after the image is enhanced; when the central response value is smaller than the first total suppression value, the value of 0 is used as an image enhancement.
  • the target pixel may be one or more pixels in an image that needs to be image enhanced.
  • image enhancement shown in the above step may be performed on other pixels in the image, that is, the image enhancement shown in the above step may be performed on the entire image.
  • the device may be any device that supports image processing, such as a computer, a mobile phone, a tablet, and the like.
  • processor 52 is further configured to perform the following operations:
  • the second total suppression value of the neighboring pixel points is a sum of suppression values of the other neighboring pixel points to the neighboring pixel points
  • the other neighboring pixel points refer to all the neighboring pixel points of the all neighboring pixel points except the neighboring pixel points;
  • the T difference of the suppression point difference of the domain pixel is the difference between the original pixel value of the neighboring pixel point minus the second total suppression value of the neighboring pixel point, Or the difference between the neighboring pixels is 0, and the T is calculated based on the original pixel value of the neighboring pixel and the distance between the neighboring pixel and the target pixel.
  • the operation performed by the processor 52 to calculate the suppression value of the target pixel point by the neighboring pixel point based on the second total suppression value of each of the neighboring pixel points may be performed.
  • a potential function of the neighboring pixel point is a norm value of a difference between an original pixel value of the neighboring pixel point and an original pixel value of the target pixel point ;
  • the rate of change of the potential function being a ratio of a power function of the potential function to a power function of all potential functions, the power function of the potential function
  • the exponent is the potential function
  • the base of the power function of the potential function is a preset base
  • a kernel value of the neighboring pixel point is a rate of change based on a potential function of the neighboring pixel point, and the neighboring pixel point and the target pixel point a value calculated from the distance value;
  • the T difference of the suppression point difference of the domain pixel is the difference between the original pixel value of the neighboring pixel point minus the second total suppression value of the neighboring pixel point, Or the difference between the suppression of the neighboring pixel points is 0, and the T is a value calculated based on the original pixel value of the neighboring pixel point and the kernel value of the neighboring pixel point.
  • the above potential function can be as follows:
  • D ( x , the potential function of the neighboring pixel point (X, y ) may be equal to 1 or equal to 2, that is, the above may be 1 norm or 2 norm, etc.
  • the above represents the neighboring pixel.
  • the pixel value of the point (X, y), and the above ( ⁇ .) represents the pixel value of the target pixel.
  • the rate of change of D ( x ) can be as follows:
  • x above can represent the rate of change of D ( x , ) of the neighborhood pixel (X , y ), and the above ex p(Z)(x, ) can represent the natural logarithm e
  • w represents a set of all neighboring pixel points, that is, the sum of the ex P( D ( x , )) of all the neighboring pixel points is obtained in the above formula.
  • the above kernel value can be calculated by the following formula: Wherein, the above represents a kernel value of a neighboring pixel point (X, y), wherein the above is a 1/H times a distance value between the neighboring pixel point (X, y) and the target pixel point, where H is a preset
  • H is a preset
  • the fixed value, for example, H can be set to 3.
  • the T in the embodiment may be a value calculated based on an original pixel value of the neighboring pixel point and a kernel value of the neighboring pixel point, and the calculation of the kernel value is based on different neighboring pixel points.
  • the interaction between the target pixel and the neighboring pixel is also combined, and the local contrast information composed of the target pixel and the neighboring pixel is used to perform a more sufficient scale on the T. Adjustment.
  • the image enhancement achieved by this embodiment in the embodiment of the present invention is defined as VSNRF.
  • the image enhancement achieved is defined as BNRF. .
  • the processor 52 subtracts the first total suppression value from the central response value to obtain a difference, and Using the difference value as a pixel value of the target pixel point after image enhancement; when the center response value is smaller than the first total suppression value, using a value of 0 as a pixel value of the target pixel point after image enhancement Operation, which can include:
  • represents the pixel value of the target pixel after the image is enhanced, and represents the calculated center response value, and represents the calculated first total suppression value.
  • the max operation described above makes the total response non-negative and implies a nonlinear effect, and is a Kroneck function.
  • the processor 52 is further configured to perform the following operations before performing an operation of calculating a center response value of the target pixel point based on the original pixel value of the target pixel point:
  • the template includes a central area, and each of the central areas is provided with a neighborhood area;
  • the target pixel point refers to a pixel point covered by the central area, and the target pixel point The neighboring pixel points the pixel points covered by the neighborhood of the central area.
  • the calculation template may be a template of 3 ⁇ 3, 5 ⁇ 5, 7 ⁇ 7, etc., wherein the number of central regions included in different templates may be different, for example, the template of 3 ⁇ 3 includes one
  • a 5 x 5 template can contain one or more central areas.
  • the size of the central area may be the same as the size of one pixel in the image, that is, the target pixel is one; the size of the central area may be the same as the size of several pixels in the image, that is, the target. There are multiple pixels.
  • the neighborhood area of the central area has an uncovered pixel point neighborhood area that does not cover a pixel point in the image where the target pixel point is located, that is, when the target pixel point is an edge pixel point of the image
  • the neighboring pixel points of the target pixel covered by the central area include:
  • the pixel points in the image covered by the neighborhood area of the central area is set to 0.
  • the operation performed by the processor 52 to calculate the central response value of the target pixel based on the original pixel value of the target pixel may include:
  • J o represents the original pixel value of the target pixel point ( ⁇ ⁇ , y0).
  • the operations performed by the processor 52 to separately calculate the second total suppression value of each of the neighboring pixel points may include:
  • the second total suppression value for each of the neighborhood points is calculated by the following formula:
  • represents the second total suppression value of the neighborhood pixel ( x , ⁇ ( x , represents the removal of the neighborhood pixel ( x , a set of all neighborhood points;
  • ⁇ , ⁇ + ") 4 (dur3) JJ (x ⁇ _ (x+m ⁇ n) represents the suppression value of the neighborhood pixel (x + m, + ") to the neighborhood pixel ( ⁇ , y), and 4 is the sensitivity coefficient between the neighborhood pixels (Example: 4 1), equal to 1/K times the distance between the farthest neighboring pixel points in the template shown in Figure 4,
  • the operation performed by the processor 52 to calculate the suppression value of the target pixel point by the neighboring pixel point based on the second total suppression value of each of the neighboring pixel points may include:
  • the suppression value of each of the neighboring pixel points to the target pixel point is calculated by the following formula:
  • denotes the neighborhood pixel point ( x , the suppression value of the target pixel point ( ⁇ , ⁇ ), 4 is the sensitivity of the neighborhood pixel point to the target pixel point, and is the neighborhood pixel point
  • is the Gaussian function parameter
  • the max operation makes the inhibition a non-negative value, implying a nonlinear effect.
  • ⁇ ) is the Kroneck function.
  • ⁇ 2 in the above formula may be as described above.
  • ⁇ 2 in the above formula may be equal to or equal to 1/H times the distance value of the target pixel point ( ⁇ ⁇ ) and the neighborhood pixel point ( x , .
  • the operation performed by the processor 52 to use the sum of the suppression values of all the neighboring pixel points and the target pixel points as the first total suppression value of the target pixel point may include:
  • the formula calculates the sum of the suppression values of all neighborhood pixels to the target pixel as the first total suppression of the target pixel:
  • N the set of neighborhood points, and there are 8 neighborhood points in this template.
  • the center of the target pixel point based on the original pixel value of the target pixel point a first response value of the target pixel point, the first total suppression value being a sum of suppression values of the target pixel, the neighborhood pixel is a neighborhood pixel previously designated as the target pixel, and the suppression value of the neighborhood pixel to the target pixel refers to the neighborhood a value represented by a suppression effect of the pixel point on the target pixel point; when the center response value is greater than the first total suppression value, subtracting the first total suppression value from the central response value to obtain a difference, And using the difference value as a pixel value of the target pixel after image enhancement; when the center response value is smaller than the first total suppression value, using a value of 0 as a pixel value of the target pixel after image enhancement .
  • the image enhancement of the target pixel is determined by the original pixel value of the target pixel and the original pixel value of the other neighboring pixel points, a simple power conversion is performed on the pixel value of the pixel only compared with the prior art.
  • the image after image enhancement is more detailed.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Landscapes

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

Abstract

本发明实施例公开了一种图像增强方法,包括:基于目标像素点的原像素值计算目标像素点的中心响应值;计算目标像素点的第一总抑制值,第一总抑制值是目标像素点所有的邻域像素点对目标像素点的抑制值的总和,邻域像素点为预先指定为目标像素点的邻域像素点,邻域像素点对目标像素点的抑制值是指该邻域像素点对目标像素点所起抑制作用所表示的值;当中心响应值大于第一总抑制值时,将中心响应值减去第一总抑制值得到差值,并将差值作为图像增强后目标像素点的像素值;当中心响应值小于第一总抑制值时,将0值作为图像增强后目标像素点的像素值。相应地,本发明实施例还公开一种图像增强设备。本发明实施例可以实现增强后的图像比较细致。

Description

一种图像增强方法及设备
本申请要求于 2013 年 4 月 23 日提交中国专利局、 申请号为 201310143064.2 、 发明名称为 "一种图像增强方法及设备" 的中国专利申请 的优先权, 其全部内容通过引用结合在本申请中。
技术领域
本发明涉及图像处理领域, 尤其涉及一种图像增强方法及设备。 背景技术
图像增强技术是图像处理领域的关键技术之一,该技术用于改善和提升原 图像质量, 甚至揭示原图像中隐藏的信息,使之更适于人类视觉系统的观测或 后续其他功能模块的处理。且图像增强技术在遥感、动态场景分析、 自动导航、 医学图像分析等领域有重要的应用。
目前图像增强技术中最常用的是通过伽马 (Gamma )校正实现图像增强, 其原理是对输入图像像素灰度釆用幂次变换, 如正式所示:
s = cr7
其中, c 为正常数, r5分别为输入图像和输出图像像素灰度。
由于上述技术中只是对输入图像进行简单的幂次变换 ,这样只能调节图像 整体的亮度, 增强后的图像比较粗糙。 发明内容
本发明实施例提供了一种图像增强方法及设备,可以实现增强后的图像比 较细致。
第一方面, 本发明实施例提供的一种图像增强方法, 包括:
基于目标像素点的原像素值计算所述目标像素点的中心响应值,所述中心 响应值大于所述原像素值;
计算所述目标像素点的第一总抑制值,所述第一总抑制值是所述目标像素 点所有的邻域像素点对所述目标像素点的抑制值的总和,所述邻域像素点为预 先指定为所述目标像素点的邻域像素点,所述邻域像素点对所述目标像素点的 抑制值是指该邻域像素点对目标像素点所起抑制作用所表示的值;
当所述中心响应值大于所述第一总抑制值时,将所述中心响应值减去所述 第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点的像素 值;
当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述 目标像素点的像素值。
第二方面, 本发明实施例提供一种图像增强设备, 包括:
第一计算模块、 第二计算模块和图像增强模块, 其中:
所述第一计算模块,用于基于目标像素点的原像素值计算所述目标像素点 的中心响应值, 所述中心响应值大于所述原像素值;
所述第二计算模块, 用于计算所述目标像素点的第一总抑制值, 所述第一 总抑制值是所述目标像素点所有的邻域像素点对所述目标像素点的抑制值的 总和, 所述邻域像素点为预先指定为所述目标像素点的邻域像素点, 所述邻域 像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素点所起抑制 作用所表示的值;
所述图像增强模块, 用于当所述中心响应值大于所述第一总抑制值时, 将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值;以及用于当所述中心响应值小于所述第一总抑制 值时, 将 0值作为图像增强后所述目标像素点的像素值。
上述技术方案中,基于目标像素点的原像素值计算所述目标像素点的中心 响应值, 所述中心响应值大于所述原像素值; 计算所述目标像素点的第一总抑 制值,所述第一总抑制值是所述目标像素点所有的邻域像素点对所述目标像素 点的抑制值的总和,所述邻域像素点为预先指定为所述目标像素点的邻域像素 点,所述邻域像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素 点所起抑制作用所表示的值; 当所述中心响应值大于所述第一总抑制值时,将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值; 当所述中心响应值小于所述第一总抑制值时,将
0值作为图像增强后所述目标像素点的像素值。 这样由于目标像素点的图像增 强决定于目标像素点的原像素值和其它邻域像素点的原像素值,从而比起现有 技术仅对像素点的像素值进行一个简单的幂次变换,本发明实施例图像增强后 的图像比较细致。 附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施 例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地, 下面描述 中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲, 在不付 出创造性劳动的前提下, 还可以根据这些附图获得其他的附图。
图 1是本发明实施例提供的一种图像增强方法的流程示意图;
图 2是本发明实施例提供的另一种图像增强方法的流程示意图;
图 3是本发明实施例提供的另一种图像增强方法的流程示意图;
图 4是本发明实施例提供的一种可选的计算模板的示意图;
图 5是本发明实施例提供的多种图像增强的效果比较示意图;
图 6是本发明实施例提供的另一个多种图像增强的效果比较示意图; 图 7是本发明实施例提供的另一个多种图像增强的效果比较示意图; 图 8是本发明实施例提供的一种图像增强设备的结构示意图;
图 9是本发明实施例提供的另一种图像增强设备的结构示意图;
图 10是本发明实施例提供的另一种图像增强设备的结构示意图; 图 11是本发明实施例提供的另一种图像增强设备的结构示意图。 具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清 楚、 完整地描述, 显然, 所描述的实施例仅仅是本发明一部分实施例, 而不是 全部的实施例。基于本发明中的实施例, 本领域普通技术人员在没有作出创造 性劳动前提下所获得的所有其他实施例, 都属于本发明保护的范围。
图 1是本发明实施例提供的一种图像增强方法的流程示意图,如图 1所示, 包括:
101、 基于目标像素点的原像素值计算所述目标像素点的中心响应值, 所 述中心响应值大于所述原像素值;
102、 计算所述目标像素点的第一总抑制值, 所述第一总抑制值是所述目 标像素点所有的邻域像素点对所述目标像素点的抑制值的总和,所述邻域像素 点为预先指定为所述目标像素点的邻域像素点,所述邻域像素点对所述目标像 素点的抑制值是指该邻域像素点对目标像素点所起抑制作用所表示的值;
103、 当所述中心响应值大于所述第一总抑制值时, 将所述中心响应值减 去所述第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点 的像素值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强 后所述目标像素点的像素值。
可选的,上述目标像素点可以是需要进行图像增强的图像中的一个或者多 个像素点。 上述步骤中仅说明对目标像素点进行图像增强, 本实施例中还可以 对上述图像中其它像素点进行上述步骤所示的图像增强,即可以实现对整个图 像进行上述步骤所示的图像增强。
可选的,所述方法还可以应用于任何支持图像处理的设备,例如:计算机、 手机、 平板电脑等设备。
上述技术方案中,基于目标像素点的原像素值计算所述目标像素点的中心 响应值, 所述中心响应值大于所述原像素值; 计算所述目标像素点的第一总抑 制值,所述第一总抑制值是所述目标像素点所有的邻域像素点对所述目标像素 点的抑制值的总和,所述邻域像素点为预先指定为所述目标像素点的邻域像素 点,所述邻域像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素 点所起抑制作用所表示的值; 当所述中心响应值大于所述第一总抑制值时,将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述目标像素点的像素值。 这样由于目标像素点的图像增 强决定于目标像素点的原像素值和其它邻域像素点的原像素值,从而比起现有 技术仅对像素点的像素值进行一个简单的幂次变换,本发明实施例图像增强后 的图像比较细致。 图 2 是本发明实施例提供的另一种图像增强方法的流程示意图, 如图 2 所示, 包括:
201、 基于目标像素点的原像素值计算所述目标像素点的中心响应值, 所 述中心响应值大于所述原像素值。
202、 分别计算每个所述邻域像素点的第二总抑制值, 所述邻域像素点的 第二总抑制值是指其它邻域像素点对该邻域像素点的抑制值的总和,所述其它 邻域像素点是指所述所有邻域像素点中除该邻域像素点之外的所有邻域像素 点。
可选的,假设目标像素点的邻域像素包括 8个,计算其中一个邻域像素点 的第二抑制值, 就是计算其它 7 个邻域像素点对该邻域像素点的抑制值的总 和, 其它 7个对该邻域像素点的抑制值的总和,也可以是理解为是其它 7个邻 域像素点对该邻域像素点的总的抑制作用所表示的值。
203、 基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述 目标像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域 像素点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像 素值减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的 抑制差值为 0, 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点与 所述目标像素点之间的距离值而计算出的值。
可选的, 步骤 202计算出每个邻域像素点的第二总抑制值, 这样步骤 203 就可以通过第二总抑制值计算出每个邻域像素点的抑制差值,再根据预先基于 该邻域像素点的原像素值,以及该邻域像素点与所述目标像素点之间的距离值 而计算出的 T, 得到每个邻域像素点对所述目标像素点的抑制值。 其中, 上述 T可以一个函数, 或者一个实数值。
作为一种可选的实施方式, 步骤 203可以包括:
计算出每个所述邻域像素点的势函数,所述邻域像素点的势函数为该邻域 像素点的原像素值减去所述目标像素点的原像素值的差值的范数值;
计算出每个所述邻域像素点的势函数的变化率 ,所述势函数的变化率是指 该势函数的幂函数与所有势函数的幂函数的比值,所述势函数的幂函数的指数 为该势函数, 所述势函数的幂函数的底数为预先设定的底数;
计算出每个所述邻域像素点的核值,所述邻域像素点的核值是基于该邻域 像素点的势函数的变化率,以及该邻域像素点与所述目标像素点之间的距离值 而计算出的值;
基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述目标 像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域像素 点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像素值 减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的抑制 差值为 0, 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点的核值 而计算出的值。
可选的, 上述势函数可以如下式所示:
D(x,y)=\\I(^-I(Xo^\\p
可选的, D(x, 表示邻域像素点(X, y)的势函数上述 p可以是等于 1或 等于 2等, 即上述可以 1范数或者 2范数等。 上述 表示邻域像素点(X, y) 的像素值, 上述 (^。)表示目标像素点的像素值。
可选的, D(x, 的变化率可以如下式所示:
k(x = expCP(x, Q)
' ∑ expCD(x, ) 其中, 上述 x, 可以表示邻域像素点 (X, y) 的 D(x, 的变化率, 上述 exp(Z)(x, )可以表示自然对数 e的 次方,上述 w表示所有邻域像素点的集 合, 即上式中对是所有邻域像素点的 exP(D(x, )进行求和。
可选的, 上述核值可以通过如下 :
Figure imgf000008_0001
其中, 上述 表示邻域像素点 (X, y) 的核值, 上述 ^为邻域像素点 (X, y) 与目标像素点之间的距离值的 1/H倍, 其中, H为预先设定的数值, 例如, H可以设定为 3。
这样该实施方式中的上述 T就可以是基于该邻域像素点的原像素值,以及 该邻域像素点的核值而计算出的值,而该核值的计算是基于不同邻域像素点之 间的相互作用, 而且也融合了目标像素点和邻域像素点之间的自适应交互作 用,利用目标像素点与邻域像素点构成的局部对比度信息,对上述 T进行更为 充分的尺度调节。本发明实施例中将该实施方式实现的图像增强定义为变尺度 非经典感受野模型 ( Variational Scalar non-classical Receptive Field, VSNRF )。 而上述将上述基于该邻域像素点的原像素值,以及该邻域像素点与所述目标像 素点之间的距离值而计算出的上述 T的实施方式中,实现的图像增强定义为基 本的非经典感受野模型 ( Basic non-classical receptive field, BNRF )。
204、 将所有邻域像素点对所述目标像素点的抑制值的总和作为所述目标 像素点的第一总抑制值。
205、 当所述中心响应值大于所述第一总抑制值时, 将所述中心响应值减 去所述第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点 的像素值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强 后所述目标像素点的像素值。
可选的, 步骤 205可以通过如下公式实现:
T = max(Tc - TN, 0) = S(Tc - TN)
其中, T表示图像增强后所述目标像素点的像素值, 表示步骤 201计算 出的中心响应值, 表示步骤 204计算出的第一总抑制值。 上述 max操作使 总响应为非负值, 并且暗含了一种非线性作用, ·)为克罗内克函数。
作为一种可选的实施方式, 如图 3所示, 在步骤 201之前, 所述方法还可 以包括:
206、 设置图像增强的计算模板, 所述模板包含中心区域, 且为每个所述 中心区域设置有邻域区域; 所述目标像素点指所述中心区域所覆盖的像素点, 所述目标像素点的邻域像素点该中心区域的邻域区域所覆盖的像素点。
可选的, 上述计算模板可以是 3 X 3、 5 x 5、 7 x 7等模板, 其中, 不同的 模板中包括的中心区域的个数可以是不同的, 例如, 3 x 3 的模板包含一个中 心区域, 5 x 5的模板可以包含一个或者多个中心区域。
可选的, 上述中心区域的大小可以是与图像中一个像素点的大小相同, 即 上述目标像素点为一个;上述中心区域的大小可以是与图像中几个像素点的大 小相同, 即上述目标像素点为多个。
可选的,当所述中心区域的邻域区域存在未覆盖所述目标像素点所在的图 像中的像素点的未覆盖像素点邻域区域时,即上述目标像素点为图像的边缘像 素点时, 该中心区域所覆盖的目标像素点的邻域像素点包括:
该中心区域的邻域区域所覆盖的所述图像中的像素点,以及与所述未覆盖 像素点邻域区域成镜像的邻域区域所覆盖的所述图像中的像素点; 或者
该中心区域的邻域区域所覆盖的所述图像中像素点。即将上述未覆盖像素 点邻域区域都设置为 0
下面以 3 X 3的模板为例进行举例说明:
3 3 的模板可以如图 4所示, 其中, 黑色的圓圈表示中心区域, 其它 8 个白色的圓圈表示邻域区域。在该模板中设置有一坐标系, 坐标系的原点为中 心区域, 即(x ,_y。), 图 4中的虚线显示其它邻域像素点对坐标值为(x, 邻域像 素点的抑制值,所有虚线的总和为邻域像素点 (x, 的第二总抑制值, 图 4中的 实线表示邻域像素点 (x, 对目标像素点 (x ,_y。)的抑制值。
在图 4所示模板中, 步骤 201可以包括:
可选的, 步骤 201可以通过如下公式计算目标像素点的中心响应值: 其 中 , 表 示 目 标 像 素 点 (X。J。) 的 中 心 响 应 值 ,
G(x,y,a) =
Figure imgf000010_0001
+y2)/2a2)为高斯函数。 (w。)表示目标像素点 ( χθ, y0)的原像素值。 4为预先设置的中心区域响应权值系数(例如: 4=4), 为 高斯函数参数, 代表高斯核函数宽度, σι等于 3 x 3的模板的中心区 (即图 4 所示的圓圈中的区域)的半径的 1/J倍, J为预先设置的数值, 例如 J为 3, 中 心区半径就为 0.5, 因此0=1/6
在图 4所示模板中, 步骤 202可以包括:
通过如下公式计算每个所述邻域像素点的第二总抑制值:
11 d 声 AG(m,n, s)
Figure imgf000010_0002
其中, ^)表示邻域像素点 (x, 的第二总抑制值, \(x, 表示除去邻域 像素点(x, 之外所有邻域点的集合; 其中, ^α , ^+")4 ( „3) ( H 表示邻域像素点(x +∞, _y + ")对邻域像素点 y)的抑制值, 4为 邻域像素点之间的敏感系数(例如: 4=1 ), 等于图 4所示的模板中最远两 邻域像素点之间的距离的 1/K倍,Κ为预先设置为数值,例如 Κ为 3 ,即 =0.94。 图 4所示的模板最远两个邻域像素点之间的距离为 因此 = 2^/3
在图 4所示模板中, 步骤 203可以包括:
通过如下公式计算每个所述邻域像素点对所述目标像素点的抑制值:
Figure imgf000011_0001
-II(x y))) 其中, ^表示邻域像素点(x, 对目标像素点(^,^)的抑制值, 4为邻 域像素点对目标像素点抑制敏感度, 为邻域像素点 的像素值。 ^为高 斯函数参数, ^等于目标像素点(x。^。)与邻域像素点(χ, 的距离值的 1/H倍, 其中, Η为预先设定的数值, 例如, Η可以设定为 3 , σ2 =0.47。 图 4所示的 模板中心点与邻域像素点的最大距离为 7Ϊ。 max操作使抑制作用为非负值, 暗含了一种非线性作用。 ·)为克罗内克函数。
可选的,当在上述 VSNRF的实现方式,上述公式中的 σ2可以为上述 。 在上述 BNRF实施方式中,上述公式中的 σ2可以为等于等于目标像素点(Χ。Ά) 与邻域像素点(x, 的距离值的 H倍。
在图 4所示模板中, 步骤 204可以包括:
通过如下公式计算将所有邻域像素点对所述目标像素点的抑制值的总和 作为所述目标像素点的第一总抑制值:
ΤΝ = ∑ 111 其中, 表示第一总抑制值, N代表邻域点的集合, 在此模板中共有 8 个邻域点。
作为一种可选的实施方式,上述方法中可以仅对灰色或者黑白图像进行图 像增强处理,还可以对彩色图像进行图像增强处理,但对于彩色图像可以是依 次对 RGB三通进行上述方法所示的图像增强处理。
作为一种可选的实施方式,本实施例还可以对本实施例实现的图像增强后 的图像与其它方法实现的图像增强后的图像进行对比。 以一张月球图像(如图 5所示)为例进行对比说明, 该图像为一张亮度分布不均匀的图像。 图 5中的 A为原图像, B为通过直方图均衡化 ( Histogram equalization, HE )技术对原 图像进行图像增强后的图像, C为通过 Retinex (—种公知的图像算法)算法 对原图像进行图像增强后的图像, D为通过 BNRF对原图像进行图像增强后的 图像, E为通过 VSNRF对原图像进行图像增强后的图像。 可以看到, VSNRF 展示了更多的细节信息, 尤其是月球的高亮度区。 如图 A箭头所指区域, 只 有 E图没有丟失。可见,在上述几种技术中 VSNRF的图像增强效果是最好的。
作为一种可选的实施方式, 所述方法还可以对图像增强后的图像进行评 价, 具体可以通过如下至少一个指标实现对图像增强后的图像的评价:
EC ( Evaluation of Contrast )指标、 EME ( Measure of Enhancement )指标 (其中, EC和 EME为图像处理领域中两个公知的指标, 没有具体的中文意 思)和明暗度保持 ( Intensity Preservation, IP )指标;
其中, 上述 EC满足如下公式:
EC(g) = log(log(∑∑ lGr (", m)2 + Gc (", m)2 )) 为上述 EC指标, 和 是横向和纵向的本实施例设置的计算模板 (例如图 4所示的 3 x 3计算模板) 中索贝尔 (Sobel )算子作用于图像增强后 的图像 (g)的结果, (n, m)是图像上的像素点。 因为 Sobel算子具有边缘检测作 用, 所以 EC衡量了算法对对边缘的图像增强作用, EC越大, 说明图像增强 的效果越好。
其中, 上述 EME满足如下公式:
Figure imgf000012_0001
, 其中, ^'和 分别为增强图像局部小块 内的最大像素值和最小像 素值。 我们釆用 8 x 8的小块, 即 = = 8。 明显, 当 ΕΜΕ越大, 表明相应增 强算法的局部对比度增强越好。 其次, 釆用 ΕΜΕ算子衡量图像增强后局部小 块的对比度增强效果:
最后, IP 指标用来避免增强后图像与原图像在整体感觉上的明显差异, 特别是避免在序列图像中的反差过大。 我们釆用 IP来表示原图像的均值与增 强后图像均值的差, 越接近于 0, 说明相应图像增强算法灰度保持的越好。 下面, 综合考虑上述三种评价指标对通过 HE技术实现的图像增强后的图 像, 通过 Retinex技术实现的图像增强后的图像, 通过 BNRF技术实现的图像 增强后的图像, 以及通过 VSNRF技术实现的图像增强后的图像进行比较, 结 果如图 6所示。
原图像为某一数据库中图像, 为如图 6最左侧一列图像所示。接着从左到 右,所示为 HE技术的结果, Retinex技术的结果, BNRF技术的结果和 VSNRF 技术的结果。 表 1是上述方法由 EC, EME以及 IP三个指标来综合衡量的结 果。 可以看到, VSNRF实现的图像增强的结果要优于其它的技术实现的图像 增强的结果。
如表 1所示, 可以客观的看到, 在上述三个指标的综合评价下, 本实施实 现的图像增强的结果要优于其它技术实现的图像增强的结果, 并且 V SNRF要 优于 BNRF。 如图 6所示, 可以主观的发现, 后两列图像表现的更为细致, 人 眼看起来增强后更为舒适。 而 HE技术有时候会导致增强后的图像一部分过亮 一部分过暗; Retinex算法有时候会导致明显的" halo"假象。
下表由 EC, EME以及 MG三个指标来综合衡量的 HE, Retinex, BNRF 和 VSNRF处理结果分别示于表 1.1 , 表 1.2和表 1.3。 可以看到, VSNRF要 优于其它的技术。
表 1.1
EC Original Image HE Retinex BNRF VSNRF
Aerial 1 2.7325 2.7222 2.7173 2.7579 2.7676
Aerial2 2.7094 2.7393 2.7537 2.7405 2.7562
Astro 1 2.6738 2.7343 2.7308 2.7219 2.7384
Barche 2.6973 2.7249 2.7262 2.7288 2.7411
Cameraman 2.6811 2.7249 2.7262 2.7128 2.7307
Clock 2.6690 2.6898 2.6944 2.6943 2.7087
Einstein 2.6758 2.7212 2.7226 2.7151 2.7308
Estatua 2.6779 2.7161 2.7348 2.7221 2.7381
Foto 2.7218 2.7408 2.7414 2.7490 2.7581
Galaxia 2.7041 2.7648 2.7626 2.7545 2.7659
Hedgebw 2.7184 2.7450 2.7633 2.7478 2.7585
Leopard 2.7131 2.7312 2.7459 2.7349 2.7472 表 1.2
Figure imgf000014_0001
表 1.3
Figure imgf000014_0002
下面还可以通过实验数据证明本实施例实现的图像增强后的图像符合人 们视觉机制。 所模拟的视错觉现象一个是马赫效应, 一个是棋盘格效应。
图 7中的 A图、 B图、 C图展示了马赫效应。 图 7 中的 A图是阶梯状的 原图, 图 7 中的 B 图是经过本发明实施例处理后的结果图。 可以看出, 图 7 中的 B 图在阶梯状边沿的暗区出现了一条更暗的线, 同样, 在亮区出现了一 个更亮的线, 这就是视错觉中的马赫现象。 图 7中的 C图是结果图的侧面图, 可以明显的看出马赫效应为突起的尖峰。
图 7 中的 D图、 E图展示了棋盘格效应。 图 7 中的 D图是具有棋盘格形 式的原图,人眼在白色的交汇处可以看到有暗的圓斑, 这就是视错觉中的棋盘 格效应。图 7中的 E图是经过本发明实施例处理后的侧面图对应于两区的剖线, 可以看到突然下陷的曲线可以解释棋盘格效应出现的圓斑错觉。
上技术方案中,在上面实施例的基础上介绍了多种可选的实施方式,且每 种实施方式都可以实现图像增强后的图像比较细致。 下面为本发明装置实施例 ,本发明装置实施例用于执行本发明方法实施例 一至二实现的方法, 为了便于说明, 仅示出了与本发明实施例相关的部分, 具 体技术细节未揭示的, 请参照本发明实施例一和实施例二。 图 8是本发明实施例提供的一种图像增强设备的结构示意图,如图 8所示, 包括: 第一计算模块 31、 第二计算模块 32和图像增强模块 33 , 其中:
第一计算模块 31 , 用于基于目标像素点的原像素值计算所述目标像素点 的中心响应值, 所述中心响应值大于所述原像素值;
第二计算模块 32 , 用于计算所述目标像素点的第一总抑制值, 所述第一 总抑制值是所述目标像素点所有的邻域像素点对所述目标像素点的抑制值的 总和, 所述邻域像素点为预先指定为所述目标像素点的邻域像素点, 所述邻域 像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素点所起抑制 作用所表示的值;
图像增强模块 33 , 用于当所述中心响应值大于所述第一总抑制值时, 将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值;以及用于当所述中心响应值小于所述第一总抑制 值时, 将 0值作为图像增强后所述目标像素点的像素值。
可选的,上述目标像素点可以是需要进行图像增强的图像中的一个或者多 个像素点。 上述步骤中仅说明对目标像素点进行图像增强, 本实施例中还可以 对上述图像中其它像素点进行上述步骤所示的图像增强,即可以实现对整个图 像进行上述步骤所示的图像增强。
可选的, 所述设备可以是任何支持图像处理的设备,例如: 计算机、手机、 平板电脑等设备。
上述技术方案中,基于目标像素点的原像素值计算所述目标像素点的中心 响应值, 所述中心响应值大于所述原像素值; 计算所述目标像素点的第一总抑 制值,所述第一总抑制值是所述目标像素点所有的邻域像素点对所述目标像素 点的抑制值的总和,所述邻域像素点为预先指定为所述目标像素点的邻域像素 点,所述邻域像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素 点所起抑制作用所表示的值; 当所述中心响应值大于所述第一总抑制值时,将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述目标像素点的像素值。 这样由于目标像素点的图像增 强决定于目标像素点的原像素值和其它邻域像素点的原像素值,从而比起现有 技术仅对像素点的像素值进行一个简单的幂次变换,本发明实施例图像增强后 的图像比较细致。 图 9 是本发明实施例提供的另一种图像增强设备的结构示意图, 如图 9 所示, 包括: 第一计算模块 41、 第二计算模块 42和图像增强模块 43 , 第二计 算模块包括第一计算子模块 421、第二计算子模块 422和第三计算子模块 423 , 其中:
第一计算模块 41 , 用于基于目标像素点的原像素值计算所述目标像素点 的中心响应值, 所述中心响应值大于所述原像素值。
第一子计算子模块 421 , 用于分别计算每个所述邻域像素点的第二总抑制 值,所述邻域像素点的第二总抑制值是指其它邻域像素点对该邻域像素点的抑 制值的总和,所述其它邻域像素点是指所述所有邻域像素点中除该邻域像素点 之外的所有邻域像素点。
可选的,假设目标像素点的邻域像素包括 8个,计算其中一个邻域像素点 的第二抑制值, 就是计算其它 7 个邻域像素点对该邻域像素点的抑制值的总 和, 其它 7个对该邻域像素点的抑制值的总和,也可以是理解为是其它 7个邻 域像素点对该邻域像素点的总的抑制作用所表示的值。
第二子计算子模块 422, 用于基于每个所述邻域像素点的第二总抑制值计 算该邻域像素点对所述目标像素点的抑制值;所述邻域像素点对所述目标像素 点的抑制值是指该邻域像素点的抑制差值的 T倍,所述邻域像素点的抑制差值 为该邻域像素点的原像素值减去该邻域像素点的第二总抑制值所得到的差值, 或者所述邻域像素点的抑制差值为 0,所述 T为预先基于该邻域像素点的原像 素值, 以及该邻域像素点与所述目标像素点之间的距离值而计算出的值。
可选的, 第一子计算子模块 421就计算每个邻域像素点的第二总抑制值, 这样第二子计算子模块 422 就可以通过第二总抑制值计算出每个邻域像素点 的抑制差值,再根据预先基于该邻域像素点的原像素值, 以及该邻域像素点与 所述目标像素点之间的距离值而计算出的 T,得到每个邻域像素点对所述目标 像素点的抑制值。 其中, 上述 T可以一个函数, 或者一个实数值。
第三子计算子模块 423 , 用于将所有邻域像素点对所述目标像素点的抑制 值的总和作为所述目标像素点的第一总抑制值。
图像增强模块 43 , 用于当所述中心响应值大于所述第一总抑制值时, 将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值;以及用于当所述中心响应值小于所述第一总抑制 值时, 将 0值作为图像增强后所述目标像素点的像素值。
可选的, 图像增强模块 43还可以用于通过如下公式实现对目标像素点的 图像增强:
T = max(Tc - TN, 0) = S(Tc - TN)
其中, T表示图像增强后所述目标像素点的像素值, 表示步骤 201计算 出的中心响应值, 表示步骤 204计算出的第一总抑制值。 上述 max操作使 总响应为非负值, 并且暗含了一种非线性作用, ·)为克罗内克函数。
作为一种可选的实施方式, 第二子计算子模块 422还可以包括: 第一计算单元(附图中未出来), 用于计算出每个所述邻域像素点的势函 数,所述邻域像素点的势函数为该邻域像素点的原像素值减去所述目标像素点 的原像素值的差值的范数值; 第二计算单元(附图中未出来), 用于计算出每个所述邻域像素点的势函 数的变化率,所述势函数的变化率是指该势函数的幂函数与所有势函数的幂函 数的比值, 所述势函数的幂函数的指数为该势函数, 所述势函数的幂函数的底 数为预先设定的底数;
第三计算单元(附图中未出来), 用于计算出每个所述邻域像素点的核值 , 所述邻域像素点的核值是基于该邻域像素点的势函数的变化率,以及该邻域像 素点与所述目标像素点之间的距离值而计算出的值;
第四计算单元(附图中未出来), 用于基于每个所述邻域像素点的第二总 抑制值计算该邻域像素点对所述目标像素点的抑制值;所述邻域像素点对所述 目标像素点的抑制值是指该邻域像素点的抑制差值的 T倍,所述邻域像素点的 抑制差值为该邻域像素点的原像素值减去该邻域像素点的第二总抑制值所得 到的差值, 或者所述邻域像素点的抑制差值为 0, 所述 T为预先基于该邻域像 素点的原像素值, 以及该邻域像素点高斯函数核而计算出的值。
可选的, 上述势函数可以如下式所示:
D(x,y) =\\ I(^ -I(Xo^ \\p
可选的, D(x, 表示邻域像素点(X, y )的势函数上术 p可以是等于 1或 等于 2等, 即上述可以 1范数或者 2范数等。 上述 表示邻域像素点(X, y ) 的像素值, 上述 (^。)表示目标像素点的像素值。
可选的, D(x, 的变化率可以如下式所示:
k(x = expCP(x, Q)
' ∑ expCD(x, ) 其中, 上述 x, 可以表示邻域像素点 (X , y ) 的 D(x, 的变化率, 上述 exp(Z)(x, )可以表示自然对数 e的 次方,上述 w表示所有邻域像素点的集 合, 即上式中对是所有邻域像素点的 exP(D(x, )进行求和。
可选的, 上述核值可以通过如下公式计算
Figure imgf000018_0001
其中, 上述 表示邻域像素点 (X , y ) 的核值, 上述 ^为邻域像素点 ( X , y ) 与目标像素点之间的距离值的 1/H倍, 其中, H为预先设定的数值, 例如, H可以设定为 3。
这样该实施方式中的上述 T就可以是基于该邻域像素点的原像素值,以及 该邻域像素点的核值而计算出的值,而该核值的计算是基于不同邻域像素点之 间的相互作用, 而且也融合了目标像素点和邻域像素点之间的自适应交互作 用,利用目标像素点与邻域像素点构成的局部对比度信息,对上述 Τ进行更为 充分的尺度调节。 本发明实施例中将该实施方式实现的图像增强定义为 VSNRF。 而上述将上述基于该邻域像素点的原像素值, 以及该邻域像素点与 所述目标像素点之间的距离值而计算出的上述 T的实施方式中,实现的图像增 强定义为基本的 BNRF。
作为一种可选的实施方式, 如图 10所示, 所述设备还可以包括: 设置单元 44, 用于设置图像增强的计算模板, 所述模板包含中心区域, 且为每个所述中心区域设置有邻域区域;所述目标像素点指所述中心区域所覆 盖的像素点,所述目标像素点的邻域像素点该中心区域的邻域区域所覆盖的像 素点。
可选的, 上述计算模板可以是 3 X 3、 5 x 5、 7 x 7等模板, 其中, 不同的 模板中包括的中心区域的个数可以是不同的, 例如, 3 x 3 的模板包含一个中 心区域, 5 x 5的模板可以包含一个或者多个中心区域。
可选的, 上述中心区域的大小可以是与图像中一个像素点的大小相同, 即 上述目标像素点为一个;上述中心区域的大小可以是与图像中几个像素点的大 小相同, 即上述目标像素点为多个。
可选的,当所述中心区域的邻域区域存在未覆盖所述目标像素点所在的图 像中的像素点的未覆盖像素点邻域区域时,该中心区域所覆盖的目标像素点的 邻域像素点包括:
该中心区域的邻域区域所覆盖的所述图像中的像素点,以及与所述未覆盖 像素点邻域区域成镜像的邻域区域所覆盖的所述图像中像素点; 或者
该中心区域的邻域区域所覆盖的所述图像中像素点。
可选的, 下面以图如图 4所示的 3 x 3的模板为例进行说明, 在该模板中 设置有一坐标系, 坐标系的原点为中心区域, 即(x。,_y。), 图 4中的虚线显示其 它邻域像素点对坐标值为(x, 邻域像素点的抑制值,所有虚线的总和为邻域像 素点 (x, 的第二总抑制值, 图 4 中的实线表示邻域像素点 (x, 对目标像素点 (x ,_y。)的抑制值。
在图 4所示模板中, 第一计算模块 41还可以通过如下公式计算目标像素 点的中心响应值:
Figure imgf000020_0001
其 中 , 表 示 目 标 像 素 点 (X。J。) 的 中 心 响 应 值 ,
Ο(χ,γ,σ) =
Figure imgf000020_0002
2)/2σ2)为高斯函数。 Jo表示目标像素点 ( χθ, y0 )的原像素值。 4为预先设置的中心区域响应权值系数(例如: 4=4 ), 为 高斯函数参数, 代表高斯核函数宽度, σι等于 3 x 3的模板的中心区 (即图 4 所示的圓圈中的区域)的半径的 1/J倍, J为预先设置的数值, 例如 J为 3, 中 心区半径就为 0.5, 因此0= 1/6
在图 4所示模板中 ,第一计算子模块 421还可以用于通过如下公式计算每 个所述邻域像素点的第二总抑制值:
11 d 声 AG(m,n, s)
Figure imgf000020_0003
其中, ^)表示邻域像素点 (x, 的第二总抑制值, \(x, 表示除去邻域 像素点(x, 之外所有邻域点的集合; 其中, ^α , ^+")4 ( „3) ( Η 表示邻域像素点 +∞, _y + ")对邻域像素点 y)的抑制值, 4为 邻域像素点之间的敏感系数(例如: 4=1 ), 等于图 4所示的模板中最远两 邻域像素点之间的距离的 1/K倍,Κ为预先设置为数值,例如 Κ为 3,即 =0.94 图 4所示的模板最远两个邻域像素点之间的距离为 因此 =2^/3
在图 4所示模板中,第二计算子模块 422通过如下公式计算每个所述邻域 像素点对所述目标像素点的抑制值:
///( ) =
Figure imgf000020_0004
= -II(xy))) 其中, ^表示邻域像素点(x, 对目标像素点(^,^)的抑制值, 4为邻 域像素点对目标像素点抑制敏感度, 为邻域像素点 的像素值。 ^为高 斯函数参数, ^等于目标像素点(x。^。)与邻域像素点(x, 的距离值的 1/H倍, 其中, H为预先设定的数值, 例如, H可以设定为 3 , σ2=0.47。 max操作使 抑制作用为非负值, 暗含了一种非线性作用。 ·)为克罗内克函数。
可选的,当在上述 VSNRF的实现方式,上述公式中的 σ2可以为上述 。 在上述 BNRF实施方式中,上述公式中的 σ2可以为等于目标像素点(X。J。)与邻 域像素点(x, 的距离值的 1/H倍。
在图 4所示模板中 ,第三计算子模块 423还可以用于通过如下公式计算将 所有邻域像素点对所述目标像素点的抑制值的总和作为所述目标像素点的第 一总抑制值:
ΤΝ = ∑ 111 其中, 表示第一总抑制值, N代表邻域点的集合, 在此模板中共有 8 个邻域点。
作为一种可选的实施方式,上述设备可以仅对灰色或者黑白图像进行图像 增强处理,还可以对彩色图像进行图像增强处理,但对于彩色图像可以是依次 对 RGB三通进行上述方法所示的图像增强处理。
上技术方案中,在上面实施例的基础上介绍了多种可选的实施方式,且每 种实施方式都可以实现图像增强后的图像比较细致。 图 11是本发明实施例提供的另一种图像增强设备的结构示意图, 如图 9 所示, 存储器 51 , 以及与存储器 51连接的处理器 52, 存储器 51用于存储一 组程序代码, 处理器 52用于调用存储器 51存储的程序执行如下操作:
基于目标像素点的原像素值计算所述目标像素点的中心响应值,所述中心 响应值大于所述原像素值;
计算所述目标像素点的第一总抑制值,所述第一总抑制值是所述目标像素 点所有的邻域像素点对所述目标像素点的抑制值的总和,所述邻域像素点为预 先指定为所述目标像素点的邻域像素点,所述邻域像素点对所述目标像素点的 抑制值是指该邻域像素点对目标像素点所起抑制作用所表示的值;
当所述中心响应值大于所述第一总抑制值时,将所述中心响应值减去所述 第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点的像素 值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述 目标像素点的像素值。
可选的,上述目标像素点可以是需要进行图像增强的图像中的一个或者多 个像素点。 上述步骤中仅说明对目标像素点进行图像增强, 本实施例中还可以 对上述图像中其它像素点进行上述步骤所示的图像增强,即可以实现对整个图 像进行上述步骤所示的图像增强。
可选的, 所述设备可以是任何支持图像处理的设备,例如: 计算机、手机、 平板电脑等设备。
在另一个可选的实施例中, 处理器 52还用于执行如下操作:
基于目标像素点的原像素值计算所述目标像素点的中心响应值,所述中心 响应值大于所述原像素值;
分别计算每个所述邻域像素点的第二总抑制值,所述邻域像素点的第二总 抑制值是指其它邻域像素点对该邻域像素点的抑制值的总和,所述其它邻域像 素点是指所述所有邻域像素点中除该邻域像素点之外的所有邻域像素点;
基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述目标 像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域像素 点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像素值 减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的抑制 差值为 0 , 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点与所述 目标像素点之间的距离值而计算出的值;
将所有邻域像素点对所述目标像素点的抑制值的总和作为所述目标像素 点的第一总抑制值;
当所述中心响应值大于所述第一总抑制值时,将所述中心响应值减去所述 第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点的像素 值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述 目标像素点的像素值。
作为一种可选的实施方式, 处理器 52执行的基于每个所述邻域像素点的 第二总抑制值计算该邻域像素点对所述目标像素点的抑制值的操作, 可以包 括:
计算出每个所述邻域像素点的势函数,所述邻域像素点的势函数为该邻域 像素点的原像素值减去所述目标像素点的原像素值的差值的范数值;
计算出每个所述邻域像素点的势函数的变化率 ,所述势函数的变化率是指 该势函数的幂函数与所有势函数的幂函数的比值,所述势函数的幂函数的指数 为该势函数, 所述势函数的幂函数的底数为预先设定的底数;
计算出每个所述邻域像素点的核值,所述邻域像素点的核值是基于该邻域 像素点的势函数的变化率,以及该邻域像素点与所述目标像素点之间的距离值 而计算出的值;
基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述目标 像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域像素 点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像素值 减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的抑制 差值为 0, 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点的核值 而计算出的值。
可选的, 上述势函数可以如下式所示:
D(x,y) =\\ I(^ -I(Xo^ \\p
可选的, D(x, 表示邻域像素点(X, y )的势函数上术 p可以是等于 1或 等于 2等, 即上述可以 1范数或者 2范数等。 上述 表示邻域像素点(X, y ) 的像素值, 上述 (^。)表示目标像素点的像素值。
可选的, D(x, 的变化率可以如下式所示:
k(x = expCP(x, Q)
' ∑ expCD(x, ) 其中, 上述 x, 可以表示邻域像素点 (X , y ) 的 D(x, 的变化率, 上述 exp(Z)(x, )可以表示自然对数 e的 次方,上述 w表示所有邻域像素点的集 合, 即上式中对是所有邻域像素点的 exP(D(x, )进行求和。
可选的, 上述核值可以通过如下公式计算得的:
Figure imgf000023_0001
其中, 上述 表示邻域像素点 (X , y ) 的核值, 上述 ^为邻域像素点 ( X , y ) 与目标像素点之间的距离值的 1/H倍, 其中, H为预先设定的数值, 例如, H可以设定为 3。
这样该实施方式中的上述 T就可以是基于该邻域像素点的原像素值,以及 该邻域像素点的核值而计算出的值,而该核值的计算是基于不同邻域像素点之 间的相互作用, 而且也融合了目标像素点和邻域像素点之间的自适应交互作 用,利用目标像素点与邻域像素点构成的局部对比度信息,对上述 T进行更为 充分的尺度调节。 本发明实施例中将该实施方式实现的图像增强定义为 VSNRF。 而上述将上述基于该邻域像素点的原像素值, 以及该邻域像素点与 所述目标像素点之间的距离值而计算出的上述 T的实施方式中,实现的图像增 强定义为 BNRF。
作为一种可选的实施方式, 处理器 52执行的当所述中心响应值大于所述 第一总抑制值时, 将所述中心响应值减去所述第一总抑制值得到差值, 并将所 述差值作为图像增强后所述目标像素点的像素值;当所述中心响应值小于所述 第一总抑制值时,将 0值作为图像增强后所述目标像素点的像素值的操作, 可 以包括:
通过如下公式实现:
T = max(Tc - TN, 0) = S(Tc - TN)
其中, τ表示图像增强后所述目标像素点的像素值, 表示计算出的中心 响应值, 表示计算出的第一总抑制值。 上述 max操作使总响应为非负值, 并且暗含了一种非线性作用, ·)为克罗内克函数。
作为一种可选的实施方式, 处理器 52在执行基于目标像素点的原像素值 计算所述目标像素点的中心响应值的操作之前, 还用于执行如下操作:
设置图像增强的计算模板, 所述模板包含中心区域,且为每个所述中心区 域设置有邻域区域; 所述目标像素点指所述中心区域所覆盖的像素点, 所述目 标像素点的邻域像素点该中心区域的邻域区域所覆盖的像素点。
可选的, 上述计算模板可以是 3 X 3、 5 x 5、 7 x 7等模板, 其中, 不同的 模板中包括的中心区域的个数可以是不同的, 例如, 3 x 3 的模板包含一个中 心区域, 5 x 5的模板可以包含一个或者多个中心区域。 可选的, 上述中心区域的大小可以是与图像中一个像素点的大小相同, 即 上述目标像素点为一个;上述中心区域的大小可以是与图像中几个像素点的大 小相同, 即上述目标像素点为多个。
可选的,当所述中心区域的邻域区域存在未覆盖所述目标像素点所在的图 像中的像素点的未覆盖像素点邻域区域时,即上述目标像素点为图像的边缘像 素点时, 该中心区域所覆盖的目标像素点的邻域像素点包括:
该中心区域的邻域区域所覆盖的所述图像中的像素点,以及与所述未覆盖 像素点邻域区域成镜像的邻域区域所覆盖的所述图像中的像素点; 或者
该中心区域的邻域区域所覆盖的所述图像中像素点。即将上述未覆盖像素 点邻域区域都设置为 0。
下面以 3 X 3的模板为例进行举例说明:
在图 4所示模板中, 处理器 52执行的基于目标像素点的原像素值计算所 述目标像素点的中心响应值的操作, 可以包括:
通过如下公式计算目标像素点的中心响应值:
Figure imgf000025_0001
其 中 , 表 示 目 标 像 素 点 (X。J。) 的 中 心 响 应 值 ,
Ο(χ,γ,σ) =
Figure imgf000025_0002
2)/2σ2)为高斯函数。 Jo表示目标像素点 ( χθ, y0)的原像素值。 4为预先设置的中心区域响应权值系数(例如: 4=4), 为 高斯函数参数, 代表高斯核函数宽度, σι等于 3x3的模板的中心区 (即图 4 所示的圓圈中的区域)的半径的 1/J倍, J为预先设置的数值, 例如 J为 3, 中 心区半径就为 0.5, 因此0=1/6
在图 4所示模板中 , 处理器 52执行的分别计算每个所述邻域像素点的第 二总抑制值的操作可以包括:
通过如下公式计算每个所述邻域 点的第二总抑制值:
11 d 声 AG(m,n, s)
Figure imgf000025_0003
其中, ^)表示邻域像素点 (x, 的第二总抑制值, \(x, 表示除去邻域 像素点(x, 之外所有邻域点的集合; 其中, ^α , ^+")4 ( „3) JJ(x^_(x+m^n)表示邻域像素点(x + m, + ")对邻域像素点(χ, y)的抑制值, 4为 邻域像素点之间的敏感系数(例如: 4=1 ), 等于图 4所示的模板中最远两 邻域像素点之间的距离的 1/K倍,Κ为预先设置为数值,例如 Κ为 3 ,即 =0.94。 图 4所示的模板最远两个邻域像素点之间的距离为 因此 = 2^/3
在图 4所示模板中 , 处理器 52执行的基于每个所述邻域像素点的第二总 抑制值计算该邻域像素点对所述目标像素点的抑制值的操作, 可以包括:
通过如下公式计算每个所述邻域像素点对所述目标像素点的抑制值:
Figure imgf000026_0001
-II(x y))) 其中, ^表示邻域像素点(x, 对目标像素点(^,^)的抑制值, 4为邻 域像素点对目标像素点抑制敏感度, 为邻域像素点 的像素值。 ^为高 斯函数参数, ^等于等于目标像素点(x。^。)与邻域像素点(χ, 的距离值的 1/H 倍, 其中, Η为预先设定的数值, 例如, Η可以设定为 3 , σ2=0.47。 max操 作使抑制作用为非负值, 暗含了一种非线性作用。 ·)为克罗内克函数。
可选的,当在上述 VSNRF的实现方式,上述公式中的 σ2可以为上述 。 在上述 BNRF实施方式中,上述公式中的 σ2可以为等于等于目标像素点(Χ。Ά) 与邻域像素点(x, 的距离值的 1/H倍。
在图 4所示模板中 , 处理器 52执行的将所有邻域像素点对所述目标像素 点的抑制值的总和作为所述目标像素点的第一总抑制值的操作, 可以包括: 通过如下公式计算将所有邻域像素点对所述目标像素点的抑制值的总和 作为所述目标像素点的第一总抑制值:
ΤΝ = ∑ 111 其中, 表示第一总抑制值, N代表邻域点的集合, 在此模板中共有 8 个邻域点。
作为一种可选的实施方式,上述方法中可以仅对灰色或者黑白图像进行图 像增强处理,还可以对彩色图像进行图像增强处理,但对于彩色图像可以是依 次对 RGB三通进行上述方法所示的图像增强处理。
上述技术方案中,基于目标像素点的原像素值计算所述目标像素点的中心 响应值, 所述中心响应值大于所述原像素值; 计算所述目标像素点的第一总抑 制值,所述第一总抑制值是所述目标像素点所有的邻域像素点对所述目标像素 点的抑制值的总和,所述邻域像素点为预先指定为所述目标像素点的邻域像素 点,所述邻域像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素 点所起抑制作用所表示的值; 当所述中心响应值大于所述第一总抑制值时,将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值; 当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述目标像素点的像素值。 这样由于目标像素点的图像增 强决定于目标像素点的原像素值和其它邻域像素点的原像素值,从而比起现有 技术仅对像素点的像素值进行一个简单的幂次变换,本发明实施例图像增强后 的图像比较细致。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程 , 是可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一计算 机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。 其中,所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory, ROM )或随机存取存储器( Random Access Memory, 简称 RAM )等。
以上所揭露的仅为本发明较佳实施例而已,当然不能以此来限定本发明之 权利范围,因此依本发明权利要求所作的等同变化,仍属本发明所涵盖的范围。

Claims

权 利 要 求
1、 一种图像增强方法, 其特征在于, 包括:
基于目标像素点的原像素值计算所述目标像素点的中心响应值,所述中心 响应值大于所述原像素值;
计算所述目标像素点的第一总抑制值,所述第一总抑制值是所述目标像素 点所有的邻域像素点对所述目标像素点的抑制值的总和,所述邻域像素点为预 先指定为所述目标像素点的邻域像素点,所述邻域像素点对所述目标像素点的 抑制值是指该邻域像素点对目标像素点所起抑制作用所表示的值;
当所述中心响应值大于所述第一总抑制值时,将所述中心响应值减去所述 第一总抑制值得到差值,并将所述差值作为图像增强后所述目标像素点的像素 值;
当所述中心响应值小于所述第一总抑制值时,将 0值作为图像增强后所述 目标像素点的像素值。
2、 如权利要求 1所述的方法, 其特征在于, 所述计算所述目标像素点的 第一总抑制值, 包括:
分别计算每个所述邻域像素点的第二总抑制值,所述邻域像素点的第二总 抑制值是指其它邻域像素点对该邻域像素点的抑制值的总和,所述其它邻域像 素点是指所述所有邻域像素点中除该邻域像素点之外的所有邻域像素点; 基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述目标 像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域像素 点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像素值 减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的抑制 差值为 0, 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点与所述 目标像素点之间的距离值而计算出的值;
将所有邻域像素点对所述目标像素点的抑制值的总和作为所述目标像素 点的第一总抑制值。
3、 如权利要求 2所述的方法, 其特征在于, 所述基于每个所述邻域像素 点的第二总抑制值计算该邻域像素点对所述目标像素点的抑制值, 包括: 计算出每个所述邻域像素点的势函数,所述邻域像素点的势函数为该邻域 像素点的原像素值减去所述目标像素点的原像素值的差值的范数值; 计算出每个所述邻域像素点的势函数的变化率 ,所述势函数的变化率是指 该势函数的幂函数与所有势函数的幂函数的比值,所述势函数的幂函数的指数 为该势函数, 所述势函数的幂函数的底数为预先设定的底数;
计算出每个所述邻域像素点的核值,所述邻域像素点的核值是基于该邻域 像素点的势函数的变化率,以及该邻域像素点与所述目标像素点之间的距离值 而计算出的值;
基于每个所述邻域像素点的第二总抑制值计算该邻域像素点对所述目标 像素点的抑制值;所述邻域像素点对所述目标像素点的抑制值是指该邻域像素 点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域像素点的原像素值 减去该邻域像素点的第二总抑制值所得到的差值,或者所述邻域像素点的抑制 差值为 0, 所述 T为基于该邻域像素点的原像素值, 以及该邻域像素点的核值 而计算出的值。
4、 如权利要求 1-3 中任一项所述的方法, 其特征在于, 所述基于目标像 素点的原像素值计算所述目标像素点的中心响应值之前, 所述方法还包括: 设置图像增强的计算模板, 所述模板包含中心区域,且为每个所述中心区 域设置有邻域区域; 所述目标像素点指所述中心区域所覆盖的像素点, 所述目 标像素点的邻域像素点该中心区域的邻域区域所覆盖的像素点。
5、 如权利要求 4所述的方法, 其特征在于, 当所述中心区域的邻域区域 存在未覆盖所述目标像素点所在的图像中的像素点的未覆盖像素点邻域区域 时, 该中心区域所覆盖的目标像素点的邻域像素点包括:
该中心区域的邻域区域所覆盖的所述图像中的像素点,以及与所述未覆盖 像素点邻域区域成镜像的邻域区域所覆盖的所述图像中的像素点; 或者
该中心区域的邻域区域所覆盖的所述图像中像素点。
6、 一种图像增强设备, 其特征在于, 包括: 第一计算模块、 第二计算模 块和图像增强模块, 其中:
所述第一计算模块,用于基于目标像素点的原像素值计算所述目标像素点 的中心响应值, 所述中心响应值大于所述原像素值;
所述第二计算模块, 用于计算所述目标像素点的第一总抑制值, 所述第一 总抑制值是所述目标像素点所有的邻域像素点对所述目标像素点的抑制值的 总和, 所述邻域像素点为预先指定为所述目标像素点的邻域像素点, 所述邻域 像素点对所述目标像素点的抑制值是指该邻域像素点对目标像素点所起抑制 作用所表示的值;
所述图像增强模块, 用于当所述中心响应值大于所述第一总抑制值时, 将 所述中心响应值减去所述第一总抑制值得到差值,并将所述差值作为图像增强 后所述目标像素点的像素值;以及用于当所述中心响应值小于所述第一总抑制 值时, 将 0值作为图像增强后所述目标像素点的像素值。
7、 如权利要求 6所述的设备, 其特征在于, 所述第二计算模块包括: 第一子计算子模块, 用于分别计算每个所述邻域像素点的第二总抑制值, 所述邻域像素点的第二总抑制值是指其它邻域像素点对该邻域像素点的抑制 值的总和,所述其它邻域像素点是指所述所有邻域像素点中除该邻域像素点之 外的所有邻域像素点;
第二子计算子模块,用于基于每个所述邻域像素点的第二总抑制值计算该 邻域像素点对所述目标像素点的抑制值;所述邻域像素点对所述目标像素点的 抑制值是指该邻域像素点的抑制差值的 T倍,所述邻域像素点的抑制差值为该 邻域像素点的原像素值减去该邻域像素点的第二总抑制值所得到的差值,或者 所述邻域像素点的抑制差值为 0 ,所述 T为预先基于该邻域像素点的原像素值, 以及该邻域像素点与所述目标像素点之间的距离值而计算出的值;
第三子计算子模块,用于将所有邻域像素点对所述目标像素点的抑制值的 总和作为所述目标像素点的第一总抑制值。
8、 如权利要求 7所述的设备, 其特征在于, 所述第二子计算子模块包括: 第一计算单元, 用于计算出每个所述邻域像素点的势函数, 所述邻域像素 点的势函数为该邻域像素点的原像素值减去所述目标像素点的原像素值的差 值的范数值;
第二计算单元, 用于计算出每个所述邻域像素点的势函数的变化率, 所述 势函数的变化率是指该势函数的幂函数与所有势函数的幂函数的比值,所述势 函数的幂函数的指数为该势函数,所述势函数的幂函数的底数为预先设定的底 数; 第三计算单元, 用于计算出每个所述邻域像素点的核值, 所述邻域像素点 的核值是基于该邻域像素点的势函数的变化率,以及该邻域像素点与所述目标 像素点之间的距离值而计算出的值;
第四计算单元,用于基于每个所述邻域像素点的第二总抑制值计算该邻域 像素点对所述目标像素点的抑制值;所述邻域像素点对所述目标像素点的抑制 值是指该邻域像素点的抑制差值的 T倍,所述邻域像素点的抑制差值为该邻域 像素点的原像素值减去该邻域像素点的第二总抑制值所得到的差值,或者所述 邻域像素点的抑制差值为 0, 所述 T为预先基于该邻域像素点的原像素值, 以 及该邻域像素点高斯函数核而计算出的值。
9、 如权利要求 6-8中任一项所述的设备, 其特征在于, 所述设备还包括: 设置单元, 用于设置图像增强的计算模板, 所述模板包含中心区域, 且为 每个所述中心区域设置有邻域区域;所述目标像素点指所述中心区域所覆盖的 像素点, 所述目标像素点的邻域像素点该中心区域的邻域区域所覆盖的像素 点。
10、 如权利要求 9所述的设备, 其特征在于, 当所述中心区域的邻域区域 存在未覆盖所述目标像素点所在的图像中的像素点的未覆盖像素点邻域区域 时, 该中心区域所覆盖的目标像素点的邻域像素点包括:
该中心区域的邻域区域所覆盖的所述图像中的像素点,以及与所述未覆盖 像素点邻域区域成镜像的邻域区域所覆盖的所述图像中像素点; 或者
该中心区域的邻域区域所覆盖的所述图像中像素点。
PCT/CN2013/089404 2013-04-23 2013-12-13 一种图像增强方法及设备 WO2014173145A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP13883013.8A EP2983130A4 (en) 2013-04-23 2013-12-13 METHOD AND DEVICE FOR IMPROVING IMAGE
US14/920,669 US9704225B2 (en) 2013-04-23 2015-10-22 Image enhancement method and device based on non-classical receptive field suppression

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201310143064.2A CN104123697B (zh) 2013-04-23 2013-04-23 一种图像增强方法及设备
CN201310143064.2 2013-04-23

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US14/920,669 Continuation US9704225B2 (en) 2013-04-23 2015-10-22 Image enhancement method and device based on non-classical receptive field suppression

Publications (1)

Publication Number Publication Date
WO2014173145A1 true WO2014173145A1 (zh) 2014-10-30

Family

ID=51769097

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2013/089404 WO2014173145A1 (zh) 2013-04-23 2013-12-13 一种图像增强方法及设备

Country Status (4)

Country Link
US (1) US9704225B2 (zh)
EP (1) EP2983130A4 (zh)
CN (1) CN104123697B (zh)
WO (1) WO2014173145A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109636755A (zh) * 2018-12-12 2019-04-16 电子科技大学 一种通过加权估计实现红外热图像增强的方法
CN110175964A (zh) * 2019-05-30 2019-08-27 大连海事大学 一种基于拉普拉斯金字塔的Retinex图像增强方法
CN114693707A (zh) * 2020-12-31 2022-07-01 北京小米移动软件有限公司 物体轮廓模板获取方法、装置、设备及存储介质

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017219962A1 (en) * 2016-06-21 2017-12-28 Zhejiang Dahua Technology Co., Ltd. Systems and methods for image processing
CN108154478B (zh) * 2016-12-02 2019-10-08 中科星图股份有限公司 一种遥感图像处理方法
CN106908791B (zh) * 2017-03-03 2020-02-21 中国科学院电子学研究所 基于全极化圆迹sar数据的输电线提取方法
CN111512341A (zh) * 2018-09-17 2020-08-07 华为技术有限公司 一种图像处理方法和装置
CN109919945B (zh) * 2019-02-01 2022-03-25 广西科技大学 基于非经典感受野非线性两侧亚单元响应的轮廓检测方法
CN114298916B (zh) * 2021-11-11 2023-04-18 电子科技大学 一种基于灰度拉伸和局部增强的X-Ray图像增强方法
CN115082438B (zh) * 2022-07-22 2022-11-25 裕钦精密拉深技术(苏州)有限公司 一种基于计算机视觉的拉深零件质检系统
CN116883270B (zh) * 2023-07-04 2024-03-22 广州医科大学附属第四医院(广州市增城区人民医院) 一种碎石手术软镜清晰化成像系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1892696A (zh) * 2005-07-08 2007-01-10 深圳迈瑞生物医疗电子股份有限公司 超声图像边缘锐化与斑点抑制方法
US20100040303A1 (en) * 2004-06-08 2010-02-18 Stmicroelectronics S.R.I. Filtering of noisy images
CN102306378A (zh) * 2011-09-14 2012-01-04 电子科技大学 一种图像增强方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7428333B2 (en) * 2004-01-23 2008-09-23 Old Dominion University Visibility improvement in color video stream
US7433086B2 (en) * 2004-09-27 2008-10-07 General Electric Company Edge detection and correcting system and method
CN101930592B (zh) * 2009-09-23 2011-10-05 电子科技大学 一种基于视觉非经典感受野模型的图像去噪方法
US8594975B2 (en) * 2010-03-04 2013-11-26 Kla-Tencor Corporation Systems and methods for wafer edge feature detection and quantification
CN102682432A (zh) 2012-05-11 2012-09-19 中国科学院半导体研究所 基于三高斯滤波的低质指纹灰度图像增强方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100040303A1 (en) * 2004-06-08 2010-02-18 Stmicroelectronics S.R.I. Filtering of noisy images
CN1892696A (zh) * 2005-07-08 2007-01-10 深圳迈瑞生物医疗电子股份有限公司 超声图像边缘锐化与斑点抑制方法
CN102306378A (zh) * 2011-09-14 2012-01-04 电子科技大学 一种图像增强方法

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109636755A (zh) * 2018-12-12 2019-04-16 电子科技大学 一种通过加权估计实现红外热图像增强的方法
CN109636755B (zh) * 2018-12-12 2023-03-28 电子科技大学 一种通过加权估计实现红外热图像增强的方法
CN110175964A (zh) * 2019-05-30 2019-08-27 大连海事大学 一种基于拉普拉斯金字塔的Retinex图像增强方法
CN114693707A (zh) * 2020-12-31 2022-07-01 北京小米移动软件有限公司 物体轮廓模板获取方法、装置、设备及存储介质
CN114693707B (zh) * 2020-12-31 2023-09-26 北京小米移动软件有限公司 物体轮廓模板获取方法、装置、设备及存储介质

Also Published As

Publication number Publication date
EP2983130A4 (en) 2016-05-18
EP2983130A1 (en) 2016-02-10
US9704225B2 (en) 2017-07-11
CN104123697A (zh) 2014-10-29
US20160042502A1 (en) 2016-02-11
CN104123697B (zh) 2017-11-17

Similar Documents

Publication Publication Date Title
WO2014173145A1 (zh) 一种图像增强方法及设备
EP3614333B1 (en) Image processing method, storage medium, and electronic apparatus
WO2021189807A1 (zh) 图像处理方法、装置、系统和电子设备
CN106157273B (zh) 生成合成图片的方法及装置
JP7175197B2 (ja) 画像処理方法および装置、記憶媒体、コンピュータ装置
US11790499B2 (en) Certificate image extraction method and terminal device
US20110280475A1 (en) Apparatus and method for generating bokeh effect in out-focusing photography
WO2018228310A1 (zh) 图像处理方法、装置及终端
WO2014169579A1 (zh) 一种色彩增强方法及装置
JP2018517993A (ja) 実時間ビデオエンハンスメント方法、端末及び非一時的コンピュータ可読記憶媒体
CN108537758B (zh) 一种基于显示器与人眼视觉特性的图像对比度增强方法
CN110827229A (zh) 一种基于纹理加权直方图均衡化的红外图像增强方法
WO2019080712A1 (zh) 视频水印生成方法、装置及终端
CN115330640B (zh) 光照贴图降噪方法、装置、设备和介质
KR101215666B1 (ko) 오브젝트 색상 보정 방법, 시스템 및 컴퓨터 프로그램 제품
US20150249779A1 (en) Smoothing of ghost maps in a ghost artifact detection method for hdr image creation
US20160300329A1 (en) Image processor and non-transitory computer readable medium
CN106651816A (zh) 数字图像的半色调处理方法和系统
US20200236270A1 (en) Systems and methods for color matching for realistic flash images
CN109658331B (zh) 图像处理方法、装置、系统及计算机存储介质
TWI678927B (zh) 動態調整影像清晰度的方法及其影像處理裝置
EP4379651A1 (en) Image processing apparatus and method of operating the same
CN117011124A (zh) 增益图的生成方法、装置、电子设备及介质
CN116894788A (zh) 图像处理方法、装置、电子设备及存储介质
JP6376673B2 (ja) 画像処理装置

Legal Events

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

Ref document number: 13883013

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

REEP Request for entry into the european phase

Ref document number: 2013883013

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 2013883013

Country of ref document: EP