CN117830142B - Video frame-by-frame denoising method and system based on intelligent recognition image processing - Google Patents
Video frame-by-frame denoising method and system based on intelligent recognition image processing Download PDFInfo
- Publication number
- CN117830142B CN117830142B CN202410251273.7A CN202410251273A CN117830142B CN 117830142 B CN117830142 B CN 117830142B CN 202410251273 A CN202410251273 A CN 202410251273A CN 117830142 B CN117830142 B CN 117830142B
- Authority
- CN
- China
- Prior art keywords
- video frame
- noise
- characteristic
- point
- frame image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000012545 processing Methods 0.000 title claims abstract description 49
- 230000009467 reduction Effects 0.000 claims abstract description 60
- 238000001914 filtration Methods 0.000 claims abstract description 24
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 238000009432 framing Methods 0.000 claims description 12
- 238000012935 Averaging Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a video frame-by-frame denoising method and system based on intelligent recognition image processing, which relate to the technical field of image processing and comprise the following steps: grouping at least one video frame image; acquiring characteristic video frame images corresponding to the average time points from the video frame image group; performing noise point identification on the characteristic video frame image; denoising the characteristic video frame image by using a mean value filtering method, and performing contour restoration on the neighborhood of the denoised characteristic noise point; judging whether the pixel point of the pre-judging video frame image at the pre-judging coordinates is a noise point or not; judging whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number; denoising the approximate processed image at the pre-judging noise point; noise reduction is performed on the non-approximation processed image at the non-approximation noise point. By arranging the approximate image acquisition module, the contour extraction module, the contour restoration module and the image noise reduction module, the noise judgment time is shortened, and the image blurring degree is reduced.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a video frame-by-frame denoising method and system based on intelligent recognition image processing.
Background
Video noise is a random variation of luminance or color information in video, typically the appearance of electronic noise. Typically generated by the sensors and circuitry of a scanner or digital camera, and possibly also by film grain or unavoidable shot noise in an ideal photodetector. Video noise is a byproduct that must exist during video capture, and gives errors and additional information to the video.
The method for removing the noise ratio of the video is that a plurality of images are obtained by framing the video, noise points of the images are reduced by adopting a mean value filtering method, but when the noise points are identified at present, all pixel points in each image are required to be judged, the time consumption is long, long waiting time can be generated when the whole video is processed, in addition, the noise reduction area of the images is blurred by adopting the mean value filtering method, and the video display effect formed by the images is affected.
Disclosure of Invention
In order to solve the technical problems, the technical scheme provides a video frame-by-frame denoising method and system based on intelligent recognition image processing, and solves the problems that in the background technology, a more common method for denoising video frames is to obtain a plurality of images, noise points of the images are denoised by adopting a mean value filtering method, all pixel points in each image need to be judged when the noise points are recognized at present, the consumed time is long, long waiting time is generated when the whole video is processed, and in addition, the noise reduction area of the images is blurred due to the mean value filtering method, so that the video display effect of image formation is affected.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a video frame-by-frame denoising method based on intelligent recognition image processing comprises the following steps:
Carrying out framing treatment on the video to obtain at least one video frame image, wherein the at least one video frame image is arranged in time sequence;
Grouping at least one video frame image to obtain at least one group of video frame image groups;
Acquiring a time point of a first frame video frame image and a time point of a last frame video frame image in a video frame image group, averaging the time point of the first frame video frame image and the time point of the last frame video frame image to obtain an average time point, and acquiring a characteristic video frame image corresponding to the average time point in the video frame image group;
performing noise point identification on the characteristic video frame image to obtain at least one characteristic noise point;
Denoising the characteristic video frame image by using a mean value filtering method, taking a pixel point with a distance from the characteristic noise point within a preset value as a neighborhood of the characteristic noise point, extracting contours from the neighborhood of the characteristic noise point, and repairing the contours of the neighborhood of the characteristic noise point after denoising to finish the noise reduction of the characteristic video frame image;
Carrying out coordinate modeling on the characteristic video frame image to obtain at least one characteristic coordinate of at least one characteristic noise point;
Acquiring at least one pre-judging video frame image in a video frame image group where the characteristic video frame image is positioned, and generating at least one pre-judging coordinate in the pre-judging video frame image, wherein the at least one pre-judging coordinate is respectively corresponding to and equal to the at least one characteristic coordinate;
Judging whether the pixel point of the pre-judging video frame image at the pre-judging coordinates is a noise point or not, if so, taking the pixel point at the pre-judging coordinates as the pre-judging noise point, and if not, not performing any processing;
Acquiring the total number of the pre-judging noise points in a single pre-judging video frame image, acquiring the total number of the characteristic noise points in the characteristic video frame image, judging whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number, if so, not performing any processing, and if not, taking the pre-judging video frame image as an approximate processing image of the characteristic video frame image;
denoising the approximate processed image at the pre-judging noise point;
And acquiring non-approximation processed images which are not used as approximation processed images in the video frame image group, carrying out noise point identification on the non-approximation processed images to obtain at least one non-approximation noise point, and carrying out noise reduction on the non-approximation processed images at the non-approximation noise point.
Preferably, the grouping the at least one video frame image to obtain at least one video frame image group includes the following steps:
acquiring a duration range of a video, and uniformly dividing the duration range to obtain at least one duration block;
And acquiring a time point of the video frame images, and summarizing the video frame images of which the time points belong to the same duration block to obtain a video frame image group.
Preferably, the step of performing noise point recognition on the feature video frame image to obtain at least one feature noise point includes the following steps:
acquiring a first characteristic region of a first pixel point in a characteristic video frame image, wherein the first characteristic region is formed by pixel points with a distance smaller than a preset distance from the first pixel point;
Calculating an average pixel value of the first feature region;
judging whether the difference between the average pixel value and the pixel value of the first pixel point is larger than a preset pixel value, if so, judging the first pixel point as a characteristic noise point, and if not, judging the first pixel point as not the characteristic noise point;
And traversing all pixel points in the characteristic video frame image by the first pixel point to obtain at least one characteristic noise point.
Preferably, the noise reduction of the characteristic video frame image by using the mean filtering method includes the following steps:
Acquiring at least one characteristic noise point, and acquiring a second characteristic region of the characteristic noise point, wherein the second characteristic region is composed of pixel points with the distance between the second characteristic region and the characteristic noise point being smaller than a preset distance;
and calculating the average pixel value of the second characteristic region, and replacing the pixel value of the characteristic noise point by using the average pixel value of the second characteristic region to finish noise reduction.
Preferably, the contour extraction of the neighborhood of the feature noise point includes the following steps:
calculating the pixel gradient of the pixel points in the neighborhood of the characteristic noise point by using a gradient formula;
averaging pixel gradients of the pixel points in the neighborhood of the characteristic noise point to obtain an average pixel gradient;
Judging whether the difference between the pixel gradient of the pixel point and the average pixel gradient is larger than a preset difference, if so, taking the pixel point as a contour point, and if not, not performing any treatment;
summarizing at least one contour point, and completing contour extraction to obtain a characteristic contour;
The gradient formula is as follows: ，
Wherein a is a pixel gradient, b is a pixel value of a pixel point, c is a pixel value average value of pixel points adjacent to the pixel point, and d is a distance between adjacent pixel points.
Preferably, the contour restoration of the neighborhood of the noise point after noise reduction includes the following steps:
Acquiring a characteristic contour in a neighborhood of a characteristic noise point before noise reduction, and acquiring a characteristic position of the characteristic contour in the neighborhood of the characteristic noise point;
And acquiring an identification position corresponding to the feature position in the neighborhood of the feature noise point after noise reduction, and replacing the pixel point at the identification position by using the feature contour.
Preferably, the modeling the coordinates of the feature video frame image to obtain at least one feature coordinate of at least one feature noise point includes the following steps:
using the left lower corner of the characteristic video frame image as an origin of coordinates, using the left side edge of the characteristic video frame image as a y axis, using the bottom side edge of the characteristic video frame image as an x axis, and establishing an xy coordinate system;
Acquiring coordinates of the feature noise points in an xy coordinate system as feature coordinates;
Generating at least one pre-determined coordinate in the pre-determined video frame image comprises the steps of:
Using the left lower corner of the pre-judging video frame image as an origin of coordinates, using the left side edge of the pre-judging video frame image as a y axis, using the bottom side edge of the pre-judging video frame image as an x axis, and establishing an xy coordinate system;
and taking the coordinates equal to the feature coordinates in the pre-judging video frame image as pre-judging coordinates.
Preferably, the denoising the approximate processed image at the pre-determined noise point comprises the following steps:
denoising the approximate processed image by using an average filtering method;
taking a pixel point with the distance from the pre-judging noise point within a preset value as a neighborhood of the pre-judging noise point, and extracting the contour of the neighborhood of the pre-judging noise point;
And performing contour restoration on the neighborhood of the noise point subjected to noise reduction and pre-judgment to finish approximate processing image noise reduction.
Preferably, the denoising the non-approximation processed image at the non-approximation noise point includes the steps of:
Denoising the non-approximation processed image by using an average filtering method;
taking the pixel points with the distance from the non-approximate noise points within a preset value as the neighborhood of the non-approximate noise points, and extracting the contour of the neighborhood of the non-approximate noise points;
and carrying out contour restoration on the neighborhood of the non-approximate noise point after noise reduction to finish the noise reduction of the non-approximate processed image.
The video frame-by-frame denoising system based on intelligent recognition image processing is used for realizing the video frame-by-frame denoising method based on intelligent recognition image processing, and comprises the following steps:
The video framing module is used for framing the video to obtain at least one video frame image;
an image grouping module that groups at least one video frame image;
The characteristic image acquisition module acquires characteristic video frame images corresponding to the average time points from the video frame image group;
The noise identification module is used for carrying out noise identification on the characteristic video frame image to obtain at least one characteristic noise, and carrying out noise identification on the non-approximate processed image to obtain at least one non-approximate noise;
the contour extraction module is used for extracting the contour of the neighborhood of the characteristic noise point;
the contour restoration module is used for restoring the contour of the neighborhood of the characteristic noise point after noise reduction;
The image noise reduction module is used for reducing noise of the characteristic video frame image by using a mean value filtering method, reducing noise of the approximate processed image at a pre-judgment noise point and reducing noise of the non-approximate processed image at a non-approximate noise point;
The approximate image acquisition module judges whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number.
Compared with the prior art, the invention has the beneficial effects that:
By arranging the image grouping module, the approximate image acquisition module, the contour extraction module, the contour restoration module and the image noise reduction module, identifying images with similar noise points, and carrying out noise reduction on the similar images at the same noise points, the method is used without carrying out noise point judgment on each pixel point in each image, so that the time for noise point judgment can be greatly shortened, the time for processing the whole video is further shortened, meanwhile, the contour near the noise point is identified, after noise reduction, the contour is restored, the contour lines in a noise-reduced fuzzy area are clear, the image blurring degree is further reduced, and the texture of the video formed by the image is improved.
Drawings
FIG. 1 is a schematic flow chart of a video frame-by-frame denoising method based on intelligent recognition image processing;
FIG. 2 is a flow chart of grouping at least one video frame image to obtain at least one group of video frame images according to the present invention;
FIG. 3 is a schematic flow chart of the feature video frame image noise identification to obtain at least one feature noise;
FIG. 4 is a schematic diagram of a process of denoising a feature video frame image using an average filtering method according to the present invention;
FIG. 5 is a schematic diagram of a flow chart of contour extraction of a neighborhood of feature noise points according to the present invention;
FIG. 6 is a schematic diagram of a flow chart of contour restoration of a neighborhood of a feature noise point after noise reduction according to the present invention;
FIG. 7 is a schematic flow chart of at least one feature coordinate process for modeling a feature video frame image to obtain at least one feature noise point according to the present invention;
FIG. 8 is a schematic diagram of a process of denoising an approximation processed image at a pre-determined noise point according to the present invention;
FIG. 9 is a schematic diagram of a process of denoising a non-approximated processed image at a non-approximated noise point in accordance with the present invention.
Detailed Description
The following description is presented to enable one of ordinary skill in the art to make and use the invention. The preferred embodiments in the following description are by way of example only and other obvious variations will occur to those skilled in the art.
Referring to fig. 1, a method for denoising video frame by frame based on intelligent recognition image processing includes:
Carrying out framing treatment on the video to obtain at least one video frame image, wherein the at least one video frame image is arranged in time sequence;
Grouping at least one video frame image to obtain at least one group of video frame image groups;
Acquiring a time point of a first frame video frame image and a time point of a last frame video frame image in a video frame image group, averaging the time point of the first frame video frame image and the time point of the last frame video frame image to obtain an average time point, and acquiring a characteristic video frame image corresponding to the average time point in the video frame image group;
performing noise point identification on the characteristic video frame image to obtain at least one characteristic noise point;
Denoising the characteristic video frame image by using a mean value filtering method, taking a pixel point with a distance from the characteristic noise point within a preset value as a neighborhood of the characteristic noise point, extracting contours from the neighborhood of the characteristic noise point, and repairing the contours of the neighborhood of the characteristic noise point after denoising to finish the noise reduction of the characteristic video frame image;
Carrying out coordinate modeling on the characteristic video frame image to obtain at least one characteristic coordinate of at least one characteristic noise point;
Acquiring at least one pre-judging video frame image in a video frame image group where the characteristic video frame image is positioned, and generating at least one pre-judging coordinate in the pre-judging video frame image, wherein the at least one pre-judging coordinate is respectively corresponding to and equal to the at least one characteristic coordinate;
Judging whether the pixel point of the pre-judging video frame image at the pre-judging coordinates is a noise point or not, if so, taking the pixel point at the pre-judging coordinates as the pre-judging noise point, and if not, not performing any processing;
Acquiring the total number of the pre-judging noise points in a single pre-judging video frame image, acquiring the total number of the characteristic noise points in the characteristic video frame image, judging whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number, if so, not performing any processing, and if not, taking the pre-judging video frame image as an approximate processing image of the characteristic video frame image;
denoising the approximate processed image at the pre-judging noise point;
And acquiring non-approximation processed images which are not used as approximation processed images in the video frame image group, carrying out noise point identification on the non-approximation processed images to obtain at least one non-approximation noise point, and carrying out noise reduction on the non-approximation processed images at the non-approximation noise point.
In the scheme, the characteristic video frame image is used as a reference image in the video frame image group, the characteristic video frame image is an image in the middle position of the video frame image group, so that the similar image in the video frame image group has the largest proportion, at least one characteristic noise point of the characteristic video frame image is obtained, coordinates of the characteristic noise point are obtained, images except the characteristic video frame image in the video frame image group are used as pre-judging video frame images, noise point judgment is carried out at the same position as the at least one characteristic noise point in the pre-judging video frame image, noise point judgment is not carried out on all pixels in the pre-judging video frame image, a large amount of judgment time can be saved, if the coincidence degree of the noise point in the pre-judging video frame image and the noise point of the characteristic video frame image is high, the pre-judging video frame image can be used as an approximate processing image of the characteristic video frame image, the noise point position used for noise reduction processing of the approximate processing image has the same position as the characteristic noise point of the characteristic video frame image, and the noise effect of the noise reduction effect can meet the requirement;
For the non-approximation image which is not used as the approximation image in the video frame image group, as the non-approximation image is a small part, noise judgment is carried out on all pixel points in the non-approximation image;
With this approach, the overall video processing time is greatly reduced.
Referring to fig. 2, grouping at least one video frame image to obtain at least one video frame image group includes the steps of:
acquiring a duration range of a video, and uniformly dividing the duration range to obtain at least one duration block;
Acquiring a time point of a video frame image, and summarizing the video frame images of which the time point belongs to the same duration block to obtain a video frame image group;
When the time length block is very small, the time interval of the video frame images in the video frame image group is small, so that the image similarity is high, the noise point coincidence degree is high, and most of the video frame images in the video frame image group can use the same noise point processing mode.
Referring to fig. 3, performing noise identification on a feature video frame image to obtain at least one feature noise includes the following steps:
acquiring a first characteristic region of a first pixel point in a characteristic video frame image, wherein the first characteristic region is formed by pixel points with a distance smaller than a preset distance from the first pixel point;
Calculating an average pixel value of the first feature region;
judging whether the difference between the average pixel value and the pixel value of the first pixel point is larger than a preset pixel value, if so, judging the first pixel point as a characteristic noise point, and if not, judging the first pixel point as not the characteristic noise point;
Traversing all pixel points in the characteristic video frame image by the first pixel points to obtain at least one characteristic noise point;
Since the image change is continuously progressive, a point having an excessive difference from the surroundings can be determined as a noise point, and noise reduction is performed thereon.
Referring to fig. 4, the noise reduction of the characteristic video frame image using the mean filtering method includes the steps of:
Acquiring at least one characteristic noise point, and acquiring a second characteristic region of the characteristic noise point, wherein the second characteristic region is composed of pixel points with the distance between the second characteristic region and the characteristic noise point being smaller than a preset distance;
Calculating an average pixel value of the second characteristic region, and replacing the pixel value of the characteristic noise point by using the average pixel value of the second characteristic region to finish noise reduction;
The mean filtering method can make the feature noise points in the feature video frame image smooth, but because the feature noise points in the feature video frame image do not appear singly but appear in an aggregated small point-shaped set, after noise reduction by the mean filtering method, blurring may be caused in certain areas, and the main reason for blurring is that the contours in certain areas are blurred, so that the boundaries of objects in the areas are not well defined, and therefore, the contours need to be repaired.
Referring to fig. 5, the contour extraction of the neighborhood of the feature noise includes the following steps:
calculating the pixel gradient of the pixel points in the neighborhood of the characteristic noise point by using a gradient formula;
averaging pixel gradients of the pixel points in the neighborhood of the characteristic noise point to obtain an average pixel gradient;
Judging whether the difference between the pixel gradient of the pixel point and the average pixel gradient is larger than a preset difference, if so, taking the pixel point as a contour point, and if not, not performing any treatment;
summarizing at least one contour point, and completing contour extraction to obtain a characteristic contour;
The gradient formula is as follows: ，
Wherein a is a pixel gradient, b is a pixel value of a pixel point, c is a pixel value average value of pixel points adjacent to the pixel point, and d is a distance between adjacent pixel points;
The extraction of the contour of the neighborhood of the characteristic noise point is performed before noise reduction, so that the contour is clear.
Referring to fig. 6, performing contour restoration on the neighborhood of the feature noise after noise reduction includes the following steps:
Acquiring a characteristic contour in a neighborhood of a characteristic noise point before noise reduction, and acquiring a characteristic position of the characteristic contour in the neighborhood of the characteristic noise point;
Acquiring an identification position corresponding to the characteristic position in the neighborhood of the characteristic noise point after noise reduction, and replacing the pixel point at the identification position by using the characteristic contour;
after noise reduction, the outline of the identification position possibly having blurring is replaced, and then the outline of the identification position can be redefined.
Referring to fig. 7, coordinate modeling is performed on a feature video frame image to obtain at least one feature coordinate of at least one feature noise point, including the steps of:
using the left lower corner of the characteristic video frame image as an origin of coordinates, using the left side edge of the characteristic video frame image as a y axis, using the bottom side edge of the characteristic video frame image as an x axis, and establishing an xy coordinate system;
Acquiring coordinates of the feature noise points in an xy coordinate system as feature coordinates;
Generating at least one pre-determined coordinate in the pre-determined video frame image comprises the steps of:
Using the left lower corner of the pre-judging video frame image as an origin of coordinates, using the left side edge of the pre-judging video frame image as a y axis, using the bottom side edge of the pre-judging video frame image as an x axis, and establishing an xy coordinate system;
and taking the coordinates equal to the feature coordinates in the pre-judging video frame image as pre-judging coordinates.
Referring to fig. 8, denoising the approximation processed image at the pre-determined noise point includes the steps of:
denoising the approximate processed image by using an average filtering method;
taking a pixel point with the distance from the pre-judging noise point within a preset value as a neighborhood of the pre-judging noise point, and extracting the contour of the neighborhood of the pre-judging noise point;
And performing contour restoration on the neighborhood of the noise point subjected to noise reduction and pre-judgment to finish approximate processing image noise reduction.
Referring to fig. 9, denoising a non-approximation processed image at a non-approximation noise point comprises the steps of:
Denoising the non-approximation processed image by using an average filtering method;
taking the pixel points with the distance from the non-approximate noise points within a preset value as the neighborhood of the non-approximate noise points, and extracting the contour of the neighborhood of the non-approximate noise points;
and carrying out contour restoration on the neighborhood of the non-approximate noise point after noise reduction to finish the noise reduction of the non-approximate processed image.
The steps employed in denoising the non-approximation processed image and the approximation processed image are similar to the steps employed in denoising the feature noise points.
The video frame-by-frame denoising system based on intelligent recognition image processing is used for realizing the video frame-by-frame denoising method based on intelligent recognition image processing, and comprises the following steps:
The video framing module is used for framing the video to obtain at least one video frame image;
an image grouping module that groups at least one video frame image;
The characteristic image acquisition module acquires characteristic video frame images corresponding to the average time points from the video frame image group;
The noise identification module is used for carrying out noise identification on the characteristic video frame image to obtain at least one characteristic noise, and carrying out noise identification on the non-approximate processed image to obtain at least one non-approximate noise;
the contour extraction module is used for extracting the contour of the neighborhood of the characteristic noise point;
the contour restoration module is used for restoring the contour of the neighborhood of the characteristic noise point after noise reduction;
The image noise reduction module is used for reducing noise of the characteristic video frame image by using a mean value filtering method, reducing noise of the approximate processed image at a pre-judgment noise point and reducing noise of the non-approximate processed image at a non-approximate noise point;
The approximate image acquisition module judges whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number.
The working process of the video frame-by-frame denoising system based on intelligent recognition image processing is as follows:
step one: the video framing module carries out framing treatment on the video to obtain at least one video frame image, and the at least one video frame image is arranged in time sequence;
Step two: the image grouping module groups at least one video frame image to obtain at least one group of video frame image groups;
Step three: the characteristic image acquisition module acquires the time point of the first frame video frame image and the time point of the last frame video frame image in the video frame image group, averages the time point of the first frame video frame image and the time point of the last frame video frame image to obtain an average time point, and acquires the characteristic video frame image corresponding to the average time point in the video frame image group;
step four: the noise identification module carries out noise identification on the characteristic video frame image to obtain at least one characteristic noise;
Step five: the image noise reduction module is used for reducing noise of the characteristic video frame image by using a mean value filtering method, pixels with the distance from the characteristic noise point within a preset value are used as the neighborhood of the characteristic noise point, the contour extraction module is used for extracting contours of the neighborhood of the characteristic noise point, and the contour restoration module is used for restoring the contours of the neighborhood of the characteristic noise point after noise reduction, so that the noise reduction of the characteristic video frame image is completed;
Step six: carrying out coordinate modeling on the characteristic video frame image to obtain at least one characteristic coordinate of at least one characteristic noise point;
step seven: acquiring at least one pre-judging video frame image in a video frame image group where the characteristic video frame image is positioned, and generating at least one pre-judging coordinate in the pre-judging video frame image, wherein the at least one pre-judging coordinate is respectively corresponding to and equal to the at least one characteristic coordinate;
step eight: the noise identification module judges whether the pixel point of the pre-judging video frame image at the pre-judging coordinates is a noise point, if so, the pixel point at the pre-judging coordinates is taken as the pre-judging noise point, and if not, no processing is carried out;
Step nine: the method comprises the steps that an approximate image acquisition module acquires the total number of pre-judging noise points in a single pre-judging video frame image, acquires the total number of characteristic noise points in a characteristic video frame image, judges whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number, does not perform any processing if yes, and takes the pre-judging video frame image as an approximate processing image of the characteristic video frame image if no;
step ten: the image noise reduction module is used for reducing noise of the approximate processed image at the pre-judging noise point;
Step eleven: acquiring non-approximation processed images which are not used as approximation processed images in the video frame image group, performing noise point identification on the non-approximation processed images by a noise point identification module to obtain at least one non-approximation noise point, and performing noise reduction on the non-approximation processed images by an image noise reduction module.
Still further, the present solution also proposes a storage medium having a computer readable program stored thereon, the computer readable program when invoked performing the above-described video frame-by-frame denoising method based on intelligent recognition image processing.
It is understood that the storage medium may be a magnetic medium, e.g., floppy disk, hard disk, magnetic tape; optical media such as DVD; or a semiconductor medium such as a solid state disk SolidStateDisk, SSD, etc.
In summary, the invention has the advantages that: by arranging the image grouping module, the approximate image acquisition module, the contour extraction module, the contour restoration module and the image noise reduction module, identifying images with similar noise points, and carrying out noise reduction on the similar images at the same noise points, the method is used without carrying out noise point judgment on each pixel point in each image, so that the time for noise point judgment can be greatly shortened, the time for processing the whole video is further shortened, meanwhile, the contour near the noise point is identified, after noise reduction, the contour is restored, the contour lines in a noise-reduced fuzzy area are clear, the image blurring degree is further reduced, and the texture of the video formed by the image is improved.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made therein without departing from the spirit and scope of the invention, which is defined by the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (10)
1. The method for denoising the video frame by frame based on intelligent recognition image processing is characterized by comprising the following steps of:
Carrying out framing treatment on the video to obtain at least one video frame image, wherein the at least one video frame image is arranged in time sequence;
Grouping at least one video frame image to obtain at least one group of video frame image groups;
Acquiring a time point of a first frame video frame image and a time point of a last frame video frame image in a video frame image group, averaging the time point of the first frame video frame image and the time point of the last frame video frame image to obtain an average time point, and acquiring a characteristic video frame image corresponding to the average time point in the video frame image group;
performing noise point identification on the characteristic video frame image to obtain at least one characteristic noise point;
Denoising the characteristic video frame image by using a mean value filtering method, taking a pixel point with a distance from the characteristic noise point within a preset value as a neighborhood of the characteristic noise point, extracting contours from the neighborhood of the characteristic noise point, and repairing the contours of the neighborhood of the characteristic noise point after denoising to finish the noise reduction of the characteristic video frame image;
Carrying out coordinate modeling on the characteristic video frame image to obtain at least one characteristic coordinate of at least one characteristic noise point;
Acquiring at least one pre-judging video frame image in a video frame image group where the characteristic video frame image is positioned, and generating at least one pre-judging coordinate in the pre-judging video frame image, wherein the at least one pre-judging coordinate is respectively corresponding to and equal to the at least one characteristic coordinate;
Judging whether the pixel point of the pre-judging video frame image at the pre-judging coordinates is a noise point or not, if so, taking the pixel point at the pre-judging coordinates as the pre-judging noise point, and if not, not performing any processing;
Acquiring the total number of the pre-judging noise points in a single pre-judging video frame image, acquiring the total number of the characteristic noise points in the characteristic video frame image, judging whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number, if so, not performing any processing, and if not, taking the pre-judging video frame image as an approximate processing image of the characteristic video frame image;
denoising the approximate processed image at the pre-judging noise point;
And acquiring non-approximation processed images which are not used as approximation processed images in the video frame image group, carrying out noise point identification on the non-approximation processed images to obtain at least one non-approximation noise point, and carrying out noise reduction on the non-approximation processed images at the non-approximation noise point.
2. The method for frame-by-frame denoising of video based on intelligent recognition image processing according to claim 1, wherein the grouping at least one video frame image to obtain at least one group of video frame images comprises the steps of:
acquiring a duration range of a video, and uniformly dividing the duration range to obtain at least one duration block;
And acquiring a time point of the video frame images, and summarizing the video frame images of which the time points belong to the same duration block to obtain a video frame image group.
3. The method for denoising video frame by frame based on intelligent recognition image processing according to claim 2, wherein the step of performing noise point recognition on the characteristic video frame image to obtain at least one characteristic noise point comprises the following steps:
acquiring a first characteristic region of a first pixel point in a characteristic video frame image, wherein the first characteristic region is formed by pixel points with a distance smaller than a preset distance from the first pixel point;
Calculating an average pixel value of the first feature region;
judging whether the difference between the average pixel value and the pixel value of the first pixel point is larger than a preset pixel value, if so, judging the first pixel point as a characteristic noise point, and if not, judging the first pixel point as not the characteristic noise point;
And traversing all pixel points in the characteristic video frame image by the first pixel point to obtain at least one characteristic noise point.
4. A method for denoising video frame by frame based on intelligent recognition image processing according to claim 3, wherein the denoising of the characteristic video frame image by using a mean value filtering method comprises the following steps:
Acquiring at least one characteristic noise point, and acquiring a second characteristic region of the characteristic noise point, wherein the second characteristic region is composed of pixel points with the distance between the second characteristic region and the characteristic noise point being smaller than a preset distance;
and calculating the average pixel value of the second characteristic region, and replacing the pixel value of the characteristic noise point by using the average pixel value of the second characteristic region to finish noise reduction.
5. The method for removing noise from video frame by frame based on intelligent recognition image processing according to claim 4, wherein the step of extracting the contour of the neighborhood of the feature noise comprises the steps of:
calculating the pixel gradient of the pixel points in the neighborhood of the characteristic noise point by using a gradient formula;
averaging pixel gradients of the pixel points in the neighborhood of the characteristic noise point to obtain an average pixel gradient;
Judging whether the difference between the pixel gradient of the pixel point and the average pixel gradient is larger than a preset difference, if so, taking the pixel point as a contour point, and if not, not performing any treatment;
summarizing at least one contour point, and completing contour extraction to obtain a characteristic contour;
The gradient formula is as follows: ，
Wherein a is a pixel gradient, b is a pixel value of a pixel point, c is a pixel value average value of pixel points adjacent to the pixel point, and d is a distance between adjacent pixel points.
6. The method for frame-by-frame denoising of video based on intelligent recognition image processing according to claim 5, wherein the contour restoration of the neighborhood of the feature noise after denoising comprises the following steps:
Acquiring a characteristic contour in a neighborhood of a characteristic noise point before noise reduction, and acquiring a characteristic position of the characteristic contour in the neighborhood of the characteristic noise point;
And acquiring an identification position corresponding to the feature position in the neighborhood of the feature noise point after noise reduction, and replacing the pixel point at the identification position by using the feature contour.
7. The method for frame-by-frame denoising of video based on intelligent recognition image processing according to claim 6, wherein the coordinate modeling of the characteristic video frame image to obtain at least one characteristic coordinate of at least one characteristic noise point comprises the following steps:
using the left lower corner of the characteristic video frame image as an origin of coordinates, using the left side edge of the characteristic video frame image as a y axis, using the bottom side edge of the characteristic video frame image as an x axis, and establishing an xy coordinate system;
Acquiring coordinates of the feature noise points in an xy coordinate system as feature coordinates;
Generating at least one pre-determined coordinate in the pre-determined video frame image comprises the steps of:
Using the left lower corner of the pre-judging video frame image as an origin of coordinates, using the left side edge of the pre-judging video frame image as a y axis, using the bottom side edge of the pre-judging video frame image as an x axis, and establishing an xy coordinate system;
and taking the coordinates equal to the feature coordinates in the pre-judging video frame image as pre-judging coordinates.
8. The method for denoising video frame by frame based on intelligent recognition image processing according to claim 7, wherein denoising the approximate processed image at the pre-determined noise point comprises the steps of:
denoising the approximate processed image by using an average filtering method;
taking a pixel point with the distance from the pre-judging noise point within a preset value as a neighborhood of the pre-judging noise point, and extracting the contour of the neighborhood of the pre-judging noise point;
And performing contour restoration on the neighborhood of the noise point subjected to noise reduction and pre-judgment to finish approximate processing image noise reduction.
9. The method for denoising video frame-by-frame based on intelligent recognition image processing according to claim 8, wherein denoising the non-approximation processed image at the non-approximation noise point comprises the steps of:
Denoising the non-approximation processed image by using an average filtering method;
taking the pixel points with the distance from the non-approximate noise points within a preset value as the neighborhood of the non-approximate noise points, and extracting the contour of the neighborhood of the non-approximate noise points;
and carrying out contour restoration on the neighborhood of the non-approximate noise point after noise reduction to finish the noise reduction of the non-approximate processed image.
10. A video frame-by-frame denoising system based on intelligent recognition image processing, for implementing the video frame-by-frame denoising method based on intelligent recognition image processing as claimed in any one of claims 1 to 9, comprising:
The video framing module is used for framing the video to obtain at least one video frame image;
an image grouping module that groups at least one video frame image;
The characteristic image acquisition module acquires characteristic video frame images corresponding to the average time points from the video frame image group;
The noise identification module is used for carrying out noise identification on the characteristic video frame image to obtain at least one characteristic noise, and carrying out noise identification on the non-approximate processed image to obtain at least one non-approximate noise;
the contour extraction module is used for extracting the contour of the neighborhood of the characteristic noise point;
the contour restoration module is used for restoring the contour of the neighborhood of the characteristic noise point after noise reduction;
The image noise reduction module is used for reducing noise of the characteristic video frame image by using a mean value filtering method, reducing noise of the approximate processed image at a pre-judgment noise point and reducing noise of the non-approximate processed image at a non-approximate noise point;
The approximate image acquisition module judges whether the difference between the total number of the pre-judging noise points and the total number of the characteristic noise points is larger than a preset number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410251273.7A CN117830142B (en) | 2024-03-06 | 2024-03-06 | Video frame-by-frame denoising method and system based on intelligent recognition image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410251273.7A CN117830142B (en) | 2024-03-06 | 2024-03-06 | Video frame-by-frame denoising method and system based on intelligent recognition image processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117830142A CN117830142A (en) | 2024-04-05 |
| CN117830142B true CN117830142B (en) | 2024-05-14 |
Family
ID=90515643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410251273.7A Active CN117830142B (en) | 2024-03-06 | 2024-03-06 | Video frame-by-frame denoising method and system based on intelligent recognition image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117830142B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118279177B (en) * | 2024-05-29 | 2024-08-16 | 西安易诺敬业电子科技有限责任公司 | Video image enhancement method for production monitoring |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102629970A (en) * | 2012-03-31 | 2012-08-08 | 广东威创视讯科技股份有限公司 | Denoising method and system for video images |
| CN106097280A (en) * | 2016-06-23 | 2016-11-09 | 浙江工业大学之江学院 | Based on normal state against the medical ultrasound image denoising method of Gauss model |
| CN107610069A (en) * | 2017-09-29 | 2018-01-19 | 西安电子科技大学 | DVS visualization video denoising methods based on shared K SVD dictionaries |
| CN115242933A (en) * | 2021-04-22 | 2022-10-25 | 广州视源电子科技股份有限公司 | Video image denoising method, device, equipment and storage medium |
-
2024
- 2024-03-06 CN CN202410251273.7A patent/CN117830142B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102629970A (en) * | 2012-03-31 | 2012-08-08 | 广东威创视讯科技股份有限公司 | Denoising method and system for video images |
| CN106097280A (en) * | 2016-06-23 | 2016-11-09 | 浙江工业大学之江学院 | Based on normal state against the medical ultrasound image denoising method of Gauss model |
| CN107610069A (en) * | 2017-09-29 | 2018-01-19 | 西安电子科技大学 | DVS visualization video denoising methods based on shared K SVD dictionaries |
| CN115242933A (en) * | 2021-04-22 | 2022-10-25 | 广州视源电子科技股份有限公司 | Video image denoising method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117830142A (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117830142B (en) | Video frame-by-frame denoising method and system based on intelligent recognition image processing | |
| CN106530246B (en) | Image defogging method and system based on dark Yu non local priori | |
| CN109754377B (en) | Multi-exposure image fusion method | |
| CN110059634B (en) | Large-scene face snapshot method | |
| CN110954553B (en) | Lens optical detection system | |
| WO2017120796A1 (en) | Pavement distress detection method and apparatus, and electronic device | |
| CN110619652A (en) | Image registration ghost elimination method based on optical flow mapping repeated area detection | |
| CN116152261B (en) | Visual inspection system for quality of printed product | |
| CN114298985B (en) | Defect detection method, device, equipment and storage medium | |
| CN112669295A (en) | Lithium battery pole piece defect detection method based on secondary threshold segmentation theory | |
| CN110648330A (en) | Defect detection method for camera glass | |
| CN114674826A (en) | Visual detection method and detection system based on cloth | |
| CN115294149A (en) | Astronomical image background extraction method and system | |
| CN111783673B (en) | Video segmentation improvement method based on OSVOS | |
| CN109426770B (en) | Iris identification method | |
| CN112767261B (en) | Non-local denoising frame for color image and video based on generalized non-convex tensor robust principal component analysis model | |
| CN112651936B (en) | Steel plate surface defect image segmentation method and system based on image local entropy | |
| CN115660937A (en) | Image processing method, image processing apparatus, electronic device, and storage medium | |
| CN115018730A (en) | Method, device, equipment and medium for removing image stripe noise | |
| Wang et al. | Removing image artifacts from scratched lens protectors | |
| CN110852228B (en) | Method and system for extracting dynamic background and detecting foreground object in monitoring video | |
| CN115187791A (en) | ORB image matching method fusing color and scale features | |
| CN109801238B (en) | Weak illumination image enhancement method based on simplified atmospheric scattering model | |
| CN114418848A (en) | Video processing method and device, storage medium and electronic equipment | |
| CN113592738A (en) | Underwater distorted image restoration method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |