CN112749298B

CN112749298B - Video cover determining method and device, electronic equipment and computer storage medium

Info

Publication number: CN112749298B
Application number: CN202010270154.8A
Authority: CN
Inventors: 欧阳晶; 高维惜; 袁易之; 司超; 叶豪
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-04-08
Filing date: 2020-04-08
Publication date: 2024-02-09
Anticipated expiration: 2040-04-08
Also published as: CN112749298A

Abstract

The invention provides a video cover determining method, a device, electronic equipment and a computer storage medium, wherein the method comprises the following steps: acquiring a video to be processed and an original cover map of the video to be processed; screening candidate picture sets of the target cover from the video to be processed and the original cover map based on at least two picture screening strategies; the target cover is determined from the candidate set of pictures based on the sharpness and/or aesthetics of each picture in the candidate set of pictures. In the scheme, the target cover of the video to be processed is determined from at least two picture screening strategies, the definition and/or the attractive degree of the pictures, so that the influence of factors of multiple dimensions on the cover quality is considered in the final target cover, the determined target cover quality is higher, and the viewing experience of a user is more met.

Description

Video cover determining method and device, electronic equipment and computer storage medium

Technical Field

The invention relates to the technical field of multimedia data processing, in particular to a method and a device for determining a video cover, electronic equipment and a computer storage medium.

Background

In the prior art, when determining a cover image of a video, it is common to detect single-dimensional information on a picture in the video, and then select the cover of the video based on the single-dimensional information, for example, whether the single-dimensional information is some information included in the image of the video, and select a picture that does not include the information from the video as the cover of the video. In the method for determining the video cover, the characteristics of the picture can only be reflected from one angle by the single-dimensional information, so that the determined cover cannot meet the requirements of a user, and further the user experience is poor.

Disclosure of Invention

The embodiment of the invention mainly aims to provide a method, a device, electronic equipment and a computer storage medium for determining a video cover.

In a first aspect, an embodiment of the present invention provides a method for determining a video cover, including:

acquiring a video to be processed and an original cover map of the video to be processed;

screening candidate picture sets of the target cover from the video to be processed and the original cover map based on at least two picture screening strategies;

The target cover is determined from the candidate set of pictures based on the sharpness and/or aesthetics of each picture in the candidate set of pictures.

In an optional embodiment of the first aspect, based on at least two picture screening policies, screening a candidate picture set of the target cover from the video to be processed and the original cover map includes:

extracting the video to be processed to obtain a first picture set and a second picture set;

screening candidate picture sets of the target cover from the first picture set, the second picture set and the original cover picture based on at least two picture screening strategies;

determining a target cover from the candidate picture set based on the sharpness and/or aesthetic degree of each picture in the candidate picture set, comprising:

determining a first alternative picture from the third picture set based on the definition of each picture in the third picture set;

determining a second alternative picture from the fourth picture set based on the aesthetic degree of each picture in the fourth picture set;

determining a target cover from the first alternative picture and/or the second alternative picture;

if the candidate picture set comprises an original cover map, the third picture set comprises a picture in a first picture set and the original cover map in the candidate picture set, and the fourth picture set comprises a picture in a second picture set and the original cover map in the candidate picture set;

If the candidate picture set does not include the original cover picture, the third picture set is a picture in the first picture set in the candidate picture set, and the fourth picture set is a picture in the second picture set in the candidate picture set.

In an optional embodiment of the first aspect, if the candidate picture set includes an original cover map, determining, based on sharpness of each picture in the third picture set, a first candidate picture from the third picture set includes:

screening out a first candidate picture based on the definition of each picture except the original cover picture in the third picture set;

screening a first candidate picture from the first candidate picture and the original cover picture based on the definition of the first candidate picture, the definition of the original cover picture and a preset picture deduction strategy;

determining a second alternative picture from the fourth picture set based on the aesthetics of the pictures in the fourth picture set, comprising:

screening out a second candidate picture based on the aesthetic degree of each picture except the original cover picture in the fourth picture set;

and screening a second candidate picture from the second candidate picture and the original cover picture based on the attractiveness of the second candidate picture, the attractiveness of the original cover picture and a preset picture withholding strategy.

In an optional embodiment of the first aspect, determining the target cover from the candidate picture set based on the sharpness and/or aesthetic degree of each picture in the candidate picture set includes:

based on the definition and/or the attractive degree of each picture in the candidate picture set, evaluating each picture in the candidate picture set respectively to obtain the initial score of each picture;

and determining the target cover from the candidate picture set based on the initial scores of the pictures.

In an optional embodiment of the first aspect, determining the target cover from the candidate picture set based on the initial score of each picture includes:

based on a preset picture withholding strategy, withholding the initial score of each picture to obtain the final score of each picture;

and determining the target cover from the candidate picture set based on the final scores of the pictures.

In an alternative embodiment of the first aspect, the picture screening policies include at least two of a pornography screening policy, an temptation screening policy, an uncomfortable picture screening policy, an advertisement screening policy, or a mosaic screening policy.

Screening the video to be processed and the original cover map based on each of at least two picture screening strategies, and taking the pictures screened by each screening strategy as candidate picture sets of the target cover;

or alternatively;

acquiring the processing sequence of each screening strategy in at least two picture screening strategies;

and sequentially screening the video to be processed and the original cover map according to the processing sequence to obtain a candidate picture set, wherein the current strategy is used for screening pictures screened by the previous strategy.

In an alternative embodiment of the first aspect, the processing order is determined by:

acquiring screening force of each screening strategy in a historical time period;

and determining the processing sequence based on the screening degree corresponding to each screening strategy, wherein the processing sequence of the screening strategy with strong screening degree is before the processing sequence of the screening strategy with weak screening degree.

In an optional embodiment of the first aspect, if the picture filtering policy includes at least one of a pornography filtering policy and an temptation filtering policy, filtering the candidate picture set of the target cover from the video to be processed and the original cover map based on the at least two picture filtering policies includes:

Determining the yellow grade of each picture in the video to be processed and the original cover picture;

deleting pictures with yellow grades exceeding the set grade from the video to be processed and the original cover map, and taking the remaining pictures after deletion as candidate picture sets;

if the picture screening policy includes at least one of an uncomfortable picture screening policy, an advertisement screening policy and a mosaic screening policy, screening a candidate picture set of a target cover from a to-be-processed video and an original cover map based on at least two picture screening policies, including:

detecting whether the video to be processed and the original cover diagram comprise setting information or not, wherein the setting information comprises at least one of uncomfortable content information, advertisement information and mosaic information;

deleting the pictures containing the setting information from the video to be processed and the original cover map, and taking the remaining pictures after deletion as candidate picture sets.

In a second aspect, the present invention provides a video cover determining apparatus, comprising:

the video acquisition module is used for acquiring the video to be processed and an original cover map of the video to be processed;

the image screening module is used for screening candidate image sets of the target cover from the video to be processed and the original cover image based on at least two image screening strategies;

And the cover determining module is used for determining the target cover from the candidate picture set based on the definition and/or the attractive degree of each picture in the candidate picture set.

In an optional embodiment of the second aspect, the picture screening module is specifically configured to, when screening the candidate picture set of the target cover from the video to be processed and the original cover map based on at least two picture screening policies:

the cover determining module is specifically configured to, when determining the target cover from the candidate picture set based on the sharpness and/or the aesthetic degree of each picture in the candidate picture set:

In an optional embodiment of the second aspect, if the candidate picture set includes an original cover map, the picture filtering module is specifically configured to, when determining the first candidate picture from the third picture set based on the sharpness of each picture in the third picture set:

the picture screening module is specifically configured to, when determining the second candidate picture from the fourth picture set based on the aesthetic degree of each picture in the fourth picture set:

In an optional embodiment of the second aspect, the cover determining module is specifically configured to, when determining the target cover from the candidate picture set based on the sharpness and/or the aesthetic degree of each picture in the candidate picture set:

In an optional embodiment of the second aspect, the cover determining module is specifically configured to, when determining the target cover from the candidate picture set based on the initial score of each picture:

In an alternative embodiment of the second aspect, the picture screening policy includes at least two of a pornography screening policy, an temptation screening policy, an uncomfortable picture screening policy, an advertisement screening policy, or a mosaic screening policy.

or alternatively;

In an alternative embodiment of the second aspect, the processing order is determined by:

In an optional embodiment of the second aspect, if the picture filtering policy includes at least one of a pornography filtering policy and an temptation filtering policy, the picture filtering module is specifically configured to, when filtering the candidate picture set of the target cover from the video to be processed and the original cover map based on at least two picture filtering policies:

if the picture screening policy includes at least one of an uncomfortable picture screening policy, an advertisement screening policy and a mosaic screening policy, the picture screening module is specifically configured to, when screening a candidate picture set of a target cover from a to-be-processed video and an original cover picture based on at least two picture screening policies:

In a third aspect, an embodiment of the present invention provides an electronic device, including a processor and a memory; the memory has stored therein readable instructions which, when loaded and executed by the processor, implement a method as shown in any of the alternative embodiments of the first aspect described above.

In a fourth aspect, embodiments of the present invention provide a computer readable storage medium having stored therein readable instructions which, when loaded and executed by a processor, implement a method as shown in any of the alternative embodiments of the first aspect described above.

The technical scheme provided by the embodiment of the invention has the beneficial effects that:

when the method, the device, the electronic equipment and the computer storage medium for determining the video cover are used for determining the target cover of the video, the video to be processed and the original cover image are screened based on at least two image screening strategies, so that the images can be screened from at least two dimensions to obtain candidate image sets, then the candidate image sets are screened through the definition and/or the attractive degree of each image, namely the images are processed from another dimension, so that the finally obtained target cover of the video to be processed considers the influence of factors of multiple dimensions on the quality of the cover, and further the quality of the determined target cover is higher and accords with the watching experience of a user.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings that are required to be used in the description of the embodiments of the present invention will be briefly described below.

FIG. 1 is a schematic flow chart of a method for determining a video cover according to an embodiment of the present invention;

FIG. 2 is a flow chart of a method for determining a video cover according to an example provided in an embodiment of the present invention;

FIG. 3 illustrates a schematic view of a video click volume effect provided in an embodiment of the present invention;

fig. 4 is a schematic diagram illustrating a synchronization processing manner of a picture filtering policy in an example provided in an embodiment of the present invention;

FIG. 5 is a schematic diagram showing a structure of a video cover determining apparatus according to an embodiment of the present invention;

fig. 6 shows a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

In order to make the objects, features and advantages of the present invention more comprehensible, the technical solutions in the embodiments of the present invention will be clearly described in conjunction with the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.

As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless expressly stated otherwise, as understood by those skilled in the art. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. The term "and/or" as used herein includes all or any element and all combination of one or more of the associated listed items.

In order to better understand and describe the schemes of the embodiments of the present invention, some technical terms related to the embodiments of the present invention are briefly described below.

Pornography: the degree of pornography information contained in the picture reflects the low custom degree of the picture to a certain extent.

Degree of temptation: the degree of the temptation information contained in the picture reflects the low custom degree of the picture to a certain extent.

Discomfort map: such as teeth, acne, snake, etc., which are uncomfortable pictures.

Advertisement diagram: pictures containing advertising information.

Mosaic: local areas in the picture that are artificially blurred.

Black and white: redundant frames around the picture, such as black edges and white edges in the picture which influence the visual effect of the user.

Pure color: the single-dot pixel ratio is too large, for example, a black screen part and a white screen part which influence the visual effect of a user in the picture.

Threshold value: also called threshold, refers to the lowest or highest value that an effect can produce.

Random number: i.e. the result of a specific random test, a comparison experiment was performed with randomly generated values according to statistical principles.

Cache: i.e. a cache memory, which allows a high-speed data exchange.

Redis: an open source is written and supported by ANSI C language, can be based on a log type and Key-Value database which can be stored and can be persistent, and can provide multiple language APIs.

Hbase: a distributed, column-oriented open source database.

PV (Page View): i.e., the page view amount, or click-through amount, reflects all of the user click-through amounts of the page.

VV (Video View): i.e., the video play amount, reflects all of the user play amounts of the video.

Offline subjective evaluation: judging each individual risk of the project by using the knowledge of expert experience and the like, giving corresponding weights, for example, a number between 0 and 10, wherein 0 represents no risk and 10 represents maximum risk, adding the weights of the risks, and then analyzing and comparing with a risk evaluation standard.

And (5) human examination: and (5) manual auditing.

In the prior art, for the content in one information interaction platform to be distributed, a plurality of processing flows, such as content security check, content classification processing, low-quality content filtering and the like, need to be performed in the middle from verification to distribution, and finally the content with better quality can be distributed to each information interaction platform, so that the cover map, the theme and the content quality of the content to be distributed are very important no matter whether the content to be distributed is image-text or video content, and can influence the exposure and the browsing amount of the distributed content.

For video, the cover map has a more important effect in publishing, because the user first sees the cover map, based on the cover map, the user can decide whether to further browse the published content, if the cover map content does not meet the viewing requirement of the user, for example, contains some information that the user does not like, the promotion of the published content can be influenced, and the attention of the user to the published content can also be influenced. Therefore, in order to allow more and more excellent contents to be recommended to the user terminal side, the exposure and browsing amount of the released contents are increased, and the viewing experience of the user is improved, and it is necessary to improve the quality of the jacket photograph from various aspects.

The original cover map of the content to be published to the platform has the following drawbacks:

1. the cover map is fuzzy, and the picture content is not clearly seen, so that the watching experience of a user is influenced.

2. The cover map contains a large amount of pornography and temptation related information, and unhealthy content can cause user dislike and influence the viewing experience of the user.

3. The cover map is not cut completely, for example, the face of a person is not complete, and a part of the picture is missing, so that the watching experience of a user is affected.

4. Advertisement information is contained in the cover map, so that the trust degree of the user on the content can be influenced.

5. The presence of a mosaic or other platform watermark in the cover map affects the viewing experience of the user.

6. The cover map has monotonous color and does not highlight theme elements, so that the watching experience of a user is influenced.

The low-quality cover diagrams not only can influence the experience of a user, but also can reduce the passing rate of human auditing, and if the cover diagrams do not pass the auditing, the contents to be released are not recommended to the user terminal, so that the cost of the manual auditing is indirectly increased.

In order to improve the pass rate of the cover sheet and the browsing amount and exposure amount of the user, optimization is required for these defects from various aspects. Aiming at the problems in the prior art, for the video to be released, the strategy of processing the pictures in the video is mainly realized in a mode of serial detection by each picture detection module, namely, after the execution of the upstream module is finished, the downstream module is triggered, so that the processing of the video pictures is realized. In addition, in the prior art, a picture selection process is generally performed on a video picture from a single dimension, for example, whether a picture contains certain information, a picture containing such information cannot be selected as a cover picture, and then a cover of a video is selected from pictures not containing such information.

The present inventors have found through analysis that the above scheme has at least the following problems:

(1) In terms of a module architecture, each module is executed in a synchronous architecture mode, each module is mutually dependent, after the execution of an upstream module is finished, a downstream module is triggered, the architecture reliability is relatively poor, and when the downstream module fails, queue blocking can occur, and time and performance are consumed.

(2) From the aspect of a picture selection strategy, the picture detection dimension involved in the existing picture selection scheme is relatively small, comparison and picture selection are basically carried out based on a factor of one dimension, and the influence of the multidimensional factor on the quality of the video cover is not considered, so that the determined cover can not well meet the watching experience of a user.

Aiming at the technical problems in the prior art, the embodiment of the invention provides a video cover determining method, which can firstly screen the video to be processed and the original cover image based on at least two image screening strategies, and then screen the candidate image set through the definition and/or the attractive degree of each image, so that the target cover of the video to be processed can be determined from multiple dimensions, the determined target cover can further meet the requirements of users, and the user experience is improved. Especially when carrying out the picture screening based on at least two kinds of picture screening strategies, can adopt concurrent processing mode, at least two kinds of picture screening strategies screen the picture simultaneously, independent each other, the influence each other does not have between each other, can improve the processing efficiency.

The following describes the technical scheme of the present invention and how the technical scheme of the present invention solves the above technical problems in detail with specific embodiments. The following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.

The execution main body of the invention can be any electronic equipment, a server, a user terminal and the like, for example, a video release platform provides a video release function for a user, the method can be executed to determine a better cover before the user submits the video to be released to the server, and then the video with the determined cover is submitted to the server, or the method can be executed by the server before the server performs auditing on the obtained videos to determine the better cover.

Fig. 1 is a schematic flow chart of a video cover determining method according to an embodiment of the present invention, where, as shown in the figure, the present invention is illustrated by taking a server as an execution body, and the method may include steps S110 to S130, where:

step S110: and acquiring the video to be processed and an original cover map of the video to be processed.

The video to be processed can be a video uploaded by a user or a video selected from a specified database, and the specific source of the video to be processed is not limited in the embodiment of the invention. Alternatively, for an information distribution platform, the video to be processed may be the video that is desired to be distributed to the platform.

The original cover chart is a default cover of the video to be processed, that is, when the video to be processed is acquired, the video to be processed is, for example, a video to be published to the information publishing platform, which is uploaded by a user, and the original cover chart of the video may be a cover selected by the user when uploading the video, for example, the cover may be a frame of picture selected from the video to be processed, or may not be a picture selected from the video to be processed, for example, a picture related to the video to be processed, or a picture of an author of the video to be processed, which is not limited in the specific content of the original cover chart in the invention, and all the specific contents are within the protection scope of the invention.

Step S120: and screening candidate picture sets of the target cover from the video to be processed and the original cover map based on at least two picture screening strategies.

The candidate picture set may include at least two pictures, where the at least two pictures may or may not include an original cover map, that is, the original cover map may be screened out or may be retained based on a picture screening policy in a process of screening the video to be processed and the original cover map, and if screened out, the candidate picture set does not include the original cover map, and if retained, the candidate picture set includes the original cover map.

In an alternative aspect of the present invention, the picture screening policies include at least two of a pornography screening policy, a temptation screening policy, a discomfort map screening policy, an advertisement screening policy, or a mosaic screening policy.

The picture screening strategy screens pictures from five dimensions of pornography, temptation, uncomfortable picture, advertisement and mosaic. The picture containing pornography information, temptation information and uncomfortable content can cause the objection of a user, and the picture containing the information can be screened out correspondingly based on a corresponding screening strategy. And the picture containing the advertisement, the mosaic and other information can influence the watching effect of the user, so that the picture containing the advertisement, the mosaic and other information can be screened out correspondingly based on a corresponding screening strategy. Therefore, the pictures can be screened out from a plurality of different dimensions, so that the pictures in the candidate picture set more meet the watching requirements of users. In addition, a specific implementation manner of filtering the video to be processed and the original cover map based on the above respective filtering strategies will be described below.

Step S130: the target cover is determined from the candidate set of pictures based on the sharpness and/or aesthetics of each picture in the candidate set of pictures.

Wherein, after determining the candidate picture set, the target cover may be determined from the candidate picture set based on the sharpness and/or aesthetic degree of each picture in the picture set. The definition can influence the visual effect of the user on the picture content, and the aesthetic degree can also influence the visual experience of the user, so that the pictures in the candidate picture set can be screened from the angles of the definition and/or the aesthetic degree, and the obtained target cover can be more in line with the visual experience of the user.

It should be appreciated that in practical applications, as an alternative implementation, when determining the target cover from the candidate picture set based on the sharpness and/or aesthetic degree of each picture in the candidate picture set, the target cover may be determined by determining the sharpness and/or aesthetic degree of each picture in the candidate picture set after obtaining the candidate picture set.

As another optional implementation manner, when determining the candidate picture set, the initial candidate picture set of the target cover may be first screened from the video to be processed and the original cover map based on the first definition and/or the first aesthetic degree of each picture in the video to be processed and the original cover map; and screening the initial candidate picture set based on at least two picture screening strategies to obtain a candidate picture set, and screening the target cover based on the determined second definition and/or second attractiveness of each picture in the candidate picture set. The second definition and/or the second aesthetic degree of each picture in the candidate picture set may correspond to the first definition and/or the first aesthetic degree of each picture in the video to be processed and the original cover map.

In the scheme of the invention, when the target cover of the video is determined, the video to be processed and the original cover image are screened based on at least two image screening strategies, so that the images can be screened from at least two dimensions to obtain candidate image sets, and then the candidate image sets are screened through the definition and/or the attractive degree of each image, namely, the images are processed from another dimension, so that the finally obtained target cover of the video to be processed considers the influence of factors of a plurality of dimensions on the quality of the cover, and the quality of the determined target cover is higher, thereby being more in line with the watching experience of users.

In addition, the video cover determined by the scheme provided by the embodiment of the invention can better meet the watching requirement of people, so that when the video can be released only by checking, the passing rate of the video content checking can be effectively improved based on the scheme, the condition that the whole video is disabled due to the unqualified video cover is avoided, the workload of manual checking is indirectly reduced, and the checking efficiency is improved. In addition, when the video cover is released, the video cover is more in line with the watching experience of the user, so that the user can be attracted more, the click rate of the user on the video is improved, and the play amount of the video is improved.

In an alternative scheme of the invention, based on at least two picture screening strategies, candidate picture sets of a target cover are screened from a video to be processed and an original cover picture, and the method comprises the following steps:

and screening candidate picture sets of the target cover from the first picture set, the second picture set and the original cover map based on at least two picture screening strategies.

The extracting of the video to be processed may be performed in a preset extracting manner, for example, a first picture set and a second picture set are randomly extracted from the video to be processed, or a uniform extracting manner is adopted to extract the first picture set and the second picture set from the video to be processed, and the uniform extracting may have multiple extracting manners, for example, one frame of picture is extracted every set number of pictures, or one frame of picture is extracted every set time. The specific implementation mode of picture extraction is not limited in the invention, and is within the protection scope of the invention.

The first picture set and the second picture set can be picture sets containing the same picture, can also be picture sets containing partial same picture, can also be two completely independent picture sets not containing the same picture, and optionally, a first set number of pictures are extracted from the video to be processed to be used as the first picture set, and then a second set number of pictures are extracted from the remaining pictures except the first picture set in the video to be processed to be used as the second picture set.

The number of pictures in the first picture set and the number of pictures in the second picture set may be the same or different, i.e., the first set number and the second set number may be the same or different. If the same, as an example, both picture sets are 20 pictures.

As an alternative, the first picture set and the second picture set may be extracted based on the definition of each picture in the video to be processed, the first picture set is extracted based on the beauty of each picture in the video to be processed, for example, the definition and the beauty of each picture in the video to be processed are determined, then a first set number of pictures with the highest definition are extracted from each picture as the first picture set, and a second set number of pictures with the highest beauty are extracted from each picture as the second picture set.

After obtaining the two picture sets, the candidate picture which can be used as the target cover can be determined from the two picture sets and the original cover map based on the definition and the aesthetic degree of the picture, respectively, it can be understood that after the first picture set, the second picture set and the original cover map are subjected to picture screening by at least two picture screening strategies, the screening result (candidate picture set) of the original cover map has two cases, and the candidate picture set comprises the original cover map or does not comprise the original cover map, based on the two cases, when the target cover is determined based on the definition and/or the aesthetic degree of each picture in the candidate picture set, the specific implementation manner is as follows:

The scheme specifically comprises the following steps: if the candidate picture set does not include the original cover picture, that is, the original cover picture is removed from the video to be processed and the original cover picture when being screened based on at least two picture screening strategies, at this time, a first candidate picture can be determined from the first picture set based on the definition of each picture in the first picture set, alternatively, the first candidate picture can be a picture with the highest definition, meanwhile, a second candidate picture can be determined from the second picture set based on the beauty of each picture in the second picture set, alternatively, the second candidate picture can be a picture with the highest beauty, and finally, the target cover is determined from the first candidate picture and/or the second candidate picture.

If the candidate picture set does not include the original cover map, that is, the original cover map is not removed from the video to be processed and the original cover map when being screened based on at least two picture screening strategies, at this time, a first candidate picture may be determined from the third picture set based on the definition of each picture in the third picture set (the first picture set and the original cover map), alternatively, the first candidate picture may be a picture with the highest definition, meanwhile, a second candidate picture may be determined from the fourth picture set based on the attractiveness of each picture in the fourth picture set (the second picture set and the original cover map), alternatively, the second candidate picture may be a picture with the highest attractiveness, and finally, the target cover is determined from the first candidate picture and/or the second candidate picture.

It can be appreciated that in practical applications, if the target cover is determined based on only the first candidate picture, the first candidate picture is the target cover, and if the target cover is determined based on only the second candidate picture, the second candidate picture is the target cover. If the target cover is determined based on the first candidate picture and the second candidate picture, that is, the target cover is determined from the first candidate picture and the second candidate picture, an alternative implementation manner is that, for example, any one of the first candidate picture and the second candidate picture can be used as the target cover based on a random number principle.

In practical application, the definition of each picture in the third picture set can be scored, each picture in the third picture set corresponds to one definition score, the highest definition score is the first alternative picture, the beauty of each picture in the fourth picture set can be scored similarly, each picture in the fourth picture set corresponds to one beauty score, and the highest beauty score is the second alternative picture.

The sharpness score of the picture can be determined based on pixel information of the picture, the attractiveness score of the picture can be determined according to the attractiveness information of the picture, the attractiveness information refers to information conforming to the aesthetic of the masses, for example, the main body of a person in the picture is incomplete, and the color blur of the picture does not conform to the aesthetic of the masses.

When the candidate picture set includes the original cover picture, determining a first candidate picture from the third picture set and determining a second candidate picture from the fourth picture set can be achieved by the following modes:

determining a first candidate picture from the third picture set based on the sharpness of each picture in the third picture set, comprising:

And screening a first candidate picture from the first candidate picture and the original cover picture based on the definition of the first candidate picture, the definition of the original cover picture and a preset picture withholding strategy.

In the first candidate picture, the second candidate picture and the original cover picture, the original cover picture can be used as a reference picture, the reference picture is respectively compared with the two candidate pictures, and the target cover is determined based on the comparison result. As an example, the original cover map is compared with the first candidate picture in definition, the determined picture is the first candidate picture, the first candidate picture is used as the first candidate picture, the original cover map is compared with the second candidate picture in attractiveness, the determined picture is the original cover map, the original cover map is used as the second candidate picture, and then the target cover is determined from the first candidate picture and the second candidate picture. Based on the above manner of determining the target cover, the target cover can be determined from the information of two different dimensions, so that the determined target cover has higher quality.

In practical application, the first candidate picture, the second candidate picture and the original cover picture can be directly compared together, the three pictures use the same information (aesthetic degree or definition) as a comparison reference, and finally the target cover is determined based on the comparison result.

The method comprises the steps of carrying out deduction processing on a first candidate picture based on the definition corresponding to the first candidate picture and an original cover picture and a preset picture deduction strategy to obtain a first deduction result, carrying out deduction processing on the original cover picture to obtain a second deduction result, and determining a first alternative picture based on the first deduction result and the second deduction result. Similarly, based on the aesthetic degree corresponding to each of the second candidate picture and the original cover picture and a preset picture deduction strategy, deduction treatment can be performed on the second candidate picture to obtain a third deduction result, deduction treatment is performed on the original cover picture to obtain a fourth deduction result, and based on the third deduction result and the fourth deduction result, a second alternative picture is determined.

In an alternative aspect of the present invention, the picture withholding policy includes a black and white withholding policy and/or a solid withholding policy.

The black-and-white edge deduction strategy refers to that the picture contains black-and-white edges, deduction is carried out, and deduction is not carried out if the picture does not contain the black-and-white edges. The black and white edges refer to redundant frames around the picture, and the visual effect of a user can be influenced. The solid (meaning of solid, i.e., single color) withholding policy refers to that the proportion of solid image areas contained in a picture is too large, for example, the proportion of black image areas in the picture to the whole picture is larger than a set proportion value, the black image areas are solid image areas, withholding is performed if the solid image areas are included, and withholding is performed if the solid image areas are not included.

In practical applications, whether the black-white edge is included in the detected image may be implemented based on a pixel value of each pixel in the image, and whether the detected image includes or is a solid-color image area may be implemented based on a related image processing algorithm such as a color value of a pixel point, which is not described herein.

The score of the picture deduction strategy corresponding to the deduction can be set based on experience, namely, the influence degree of the black and white edges and the solid-color image area on the video playing effect can be determined, and the influence degree of the video playing effect can be measured through the PV value and/or the VV value. For example, the playing effect is measured by the PV value, the larger the PV value is, the better the video playing effect is, the less the score of the corresponding buckle is, and conversely, the smaller the PV value is, the less the playing effect is, the more the score of the corresponding buckle is.

As an example, taking a PV value to determine a deduction score as an example, taking a picture containing black and white edges as a cover of a video a, taking a picture containing a solid-color image area as a cover of the video a, playing the video a at the same time, counting corresponding PV values after playing the video containing black and white edges for a period of time, and comparing the two PV values, wherein the larger the PV value is, the smaller the corresponding deduction score is.

In practical applications, the score may be adjusted based on the PV value and/or the VV value for each set period of time, so that the determined target cover better conforms to the viewing experience of the user.

As an alternative, the score of the black-white deduction strategy corresponding to the deduction is 3 points, and the score of the pure-color deduction strategy corresponding to the deduction is 3 points.

It should be noted that, when the candidate image set does not include the original cover image, the determining of the target cover may also be implemented based on the above-mentioned deduction policy based on the first candidate image and the second candidate image, that is, the first candidate image and the second candidate image are deducted based on the deduction policy, and the image with the highest score after deduction is used as the target cover.

In an alternative aspect of the present invention, selecting one of the first picture and the second picture as the target cover may also be implemented by: and selecting one of the first alternative picture and the second alternative picture as the target cover according to a selection strategy corresponding to the current time.

The selection policy may be set based on experience, and as an example, since in the solution of the present invention, the target cover is finally selected from the sharpness picture and the aesthetic picture, which picture is selected as the target cover may be determined according to the influence degree of the sharpness picture and the aesthetic picture on the video playing effect, and the influence degree of the video playing effect may be measured by the PV value and/or the VV value. For example, the larger the VV value is, the better the video playing effect is indicated, if the cover of the video played correspondingly is a resolution picture when the VV value is highest, the corresponding selection policy in the set period may be that the resolution pictures in the first candidate picture and the second candidate picture are used as target covers, and since the VV value and the VV value are changed in real time, the PV value and the VV value in a period of time may be used as references for measuring the playing effect. Such as an average of PV values over the course of a day.

In addition, in practical applications, considering that the viewing experience of the user on the cover of the video is changed in real time, the selection policy is not always fixed and needs to be changed based on the playing effect of the video, for example, the selection policy is updated once based on the PV value and/or the VV value in a set period of time. When the scheme of the invention determines the target cover, the selection strategy corresponding to the time period is selected according to the time period of the current time.

In an alternative scheme, when the video playing effect is measured based on the PV value and the VV value, an integrated value may be obtained based on the weight corresponding to the PV value, the weight corresponding to the VV value, and the PV value and the VV value, and the video playing effect is measured based on the magnitude of the integrated value, where the larger the integrated value is, the better the playing effect is. The video playing effect is comprehensively measured from the aspects of PV and VV, so that the target cover determined based on the determined selection strategy is more in line with the viewing experience of the user.

It should be noted that, when the candidate image set does not include the original cover image, determining the target cover based on the first candidate image and the second candidate image may also be implemented by the above implementation manner, which is not described herein in detail.

In the alternative scheme of the invention, based on the five picture screening strategies described above, the specific implementation manner of screening the video to be processed and the original cover map is as follows:

if the picture screening policy includes at least one of a pornography screening policy and an temptation screening policy, screening a candidate picture set of the target cover from the to-be-processed video and the original cover map based on the at least two picture screening policies, including:

And deleting the pictures with yellow grades exceeding the set grade from the video to be processed and the original cover map, and taking the remaining pictures after deletion as candidate picture sets.

If the at least two picture screening strategies include at least one of a pornography screening strategy and an temptation screening strategy, the yellow level of each picture in the to-be-processed video and the original cover picture can be determined first, the yellow level includes the pornography and/or temptation of the picture, the picture with the yellow level exceeding the set level (the set pornography and/or the set temptation) is deleted from the to-be-processed video and the original cover picture, and the remaining pictures after deletion are used as candidate picture sets.

The pornography refers to the degree of pornography information contained in the picture, and one determination mode of the pornography can be as follows: the pornography is determined based on pornography information contained in each picture. The pornography information comprises yellow information such as pornography information of men and women, and the pornography information can be determined according to the picture content in the picture. Since pornography information can cause the user's dislike, when picture screening is performed based on a pornography screening strategy, pictures with pornography greater than the set pornography (set level) can be deleted, and pictures with pornography not greater than the set pornography can be reserved as candidate picture sets. Or, the pornography score of each picture can be obtained according to the pornography information contained in each picture, and the higher the pornography score is, the higher the pornography of the picture is.

Based on the pornography scores of the pictures, a candidate picture set may be determined by: the pornography scores in a certain range can be divided into the same pornography grade, for example, the pornography scores are 0-100, each 20 scores are divided into 1 pornography grade, the picture with the pornography grade larger than the set pornography grade is deleted, and the picture with the pornography grade not larger than the set pornography grade is reserved as a candidate picture set. Or deleting the picture with the pornography score larger than the second set threshold value, and reserving the picture with the pornography score not larger than the second set threshold value as a candidate picture set.

The temptation degree refers to the degree of containing temptation information in the picture, and one determination mode of the temptation degree may be: the temptation is determined based on the temptation information contained in each picture. The temptation information includes naked art photo, yellow information with temptation degree such as dew point of the main body, etc., and the temptation information can be determined according to the picture content in the picture, for example, the color of the main body (used for judging whether the main body is the naked art photo), the integrity (used for judging whether the dew point information exists), etc. Because the temptation information can cause the objection of the user, when the picture screening is performed based on the temptation screening policy, the picture with the temptation degree larger than the set temptation degree (the set level) can be deleted, and the picture with the temptation degree not larger than the set temptation degree can be reserved as the candidate picture set. Alternatively, the temptation degree score may be obtained for each picture according to the temptation information included in each picture, where the higher the temptation degree score, the higher the temptation degree score indicates the temptation degree of the picture.

Based on the temptation scores of the pictures, a candidate picture set may be determined by: the certain range of temptation scores may be divided into the same temptation grade, for example, the temptation scores are 0-100 points, each 20 points are divided into 1 temptation grade, the pictures with the temptation grade greater than the set temptation grade are deleted, and the pictures with the temptation grade not greater than the set temptation grade are reserved as candidate picture sets. Or deleting the pictures with the temptation scores larger than the first set threshold value, and reserving the pictures with the temptation scores not larger than the first set threshold value as candidate picture sets.

In the scheme of the invention, if the picture screening strategy comprises at least one of an uncomfortable picture screening strategy, an advertisement screening strategy and a mosaic screening strategy, the candidate picture set of the target cover is screened from the video to be processed and the original cover picture based on at least two picture screening strategies, and the method comprises the following steps:

If the at least two picture screening policies include at least one of an uncomfortable picture screening policy, an advertisement screening policy and a mosaic screening policy, whether each picture in the to-be-processed video and the original cover picture includes setting information or not may be determined first, the setting information corresponding to the uncomfortable picture screening policy is uncomfortable content information, the setting information corresponding to the advertisement screening policy is advertisement information, and the setting information corresponding to the mosaic screening policy is mosaic information. And deleting the pictures containing the setting information from the video to be processed and the original cover map, and taking the remaining pictures after deletion as candidate picture sets.

Wherein, the discomfort graph refers to a picture containing picture content causing sensory discomfort of a user, and the picture content causing sensory discomfort of the user can comprise nausea teeth, hair, acnes and the like, and the information can be called discomfort content information. The uncomfortable content information causes a sense reaction of discomfort of a user due to the uncomfortable picture information, so that when picture screening is performed based on the uncomfortable picture screening strategy, pictures containing the uncomfortable content information can be deleted, and pictures which do not contain the uncomfortable content information are reserved to be used as candidate picture sets.

The principle of the discomfort graph screening strategy is the same as that of the discomfort graph screening strategy, and because the pictures containing the advertisement information also influence the watching experience of the user, the corresponding implementation mode of the advertisement screening strategy is to delete the pictures containing the advertisement information and keep the pictures not containing the advertisement information as candidate picture sets. The advertisement information may be determined based on advertisement identification in the picture, such as advertisement title, advertisement link, etc.

The method is similar to the above-mentioned uncomfortable diagram screening policy principle, and because the mosaics in the pictures also affect the viewing experience of the user, the implementation mode corresponding to the mosaic screening policy is to delete the pictures containing mosaic information, and keep the pictures not containing mosaic information as candidate picture sets. The mosaic information may be determined based on pixel information of the picture. In practical application, the area occupation ratio of the area corresponding to the mosaic information can be determined, then the picture with the area occupation ratio exceeding a certain threshold value is deleted, and the picture with the area occupation ratio not exceeding the certain threshold value is reserved.

In an alternative of the present invention, the processing results of the video to be processed and the original cover map may be stored in a specified database. The processing results of the video to be processed and the original cover diagram comprise definition detection results, attractive degree detection results and pornography detection results of the pictures, and temptation detection results of the pictures comprise at least one of detection results of uncomfortable content information, advertisement information, mosaic information, black and white edges and solid-color image areas. The specified database may be at least one of a hbase database, a cache memory, or a redis database. The processing results are stored in the appointed database, so that when the same picture is encountered later, the corresponding processing results can be directly obtained from the appointed database, the picture does not need to be processed once again, processing steps are saved, and efficiency is improved.

In the alternative scheme of the invention, the video to be processed and the original cover map are screened based on at least two picture screening strategies by the following two modes:

firstly, the mode of concurrent processing (also called asynchronous processing) of each filtering strategy is that each filtering strategy of at least two kinds of picture filtering strategies is used for filtering the video to be processed and the original cover map respectively, and the pictures screened by each filtering strategy are used as candidate picture sets of the target cover.

The concurrent processing means that several filtering strategies can perform picture filtering processing at the same time, the input of each filtering strategy is the video to be processed and the original cover map, for example, two filtering strategies are provided, namely an uncomfortable map filtering strategy and an advertisement filtering strategy, the video to be processed and the original cover map can be screened at the same time based on the uncomfortable map filtering strategy and the advertisement filtering strategy, the two filtering strategies are not related to each other and are not influenced by each other, so that the data processing efficiency can be improved by the concurrent processing mode, and meanwhile, when another new filtering strategy is added, the processing branch of the new filtering strategy is directly added, the business logic of other filtering strategies is not influenced, and the maintenance is convenient.

Secondly, the synchronous processing mode of each screening strategy is that the processing sequence of each screening strategy in at least two picture screening strategies is obtained;

The synchronous processing mode refers to that all the screening strategies are serial, the current screening strategy is input into a picture screened by the previous screening strategy according to the processing sequence of all the screening strategies, and the first screening strategy is input into a video to be processed and an original cover picture. As an example, there are two filtering strategies, i.e., an uncomfortable image filtering strategy and an advertisement filtering strategy, and the processing sequence of the uncomfortable image filtering strategy is before the advertisement filtering strategy, the video to be processed and the original cover image are filtered based on the uncomfortable image filtering strategy to obtain a third picture set, and then the third picture set is filtered based on the advertisement filtering strategy, so that the filtered pictures are candidate picture sets.

In the concurrent processing mode, the input of each advertisement screening strategy is the same number of pictures, and the comparison consumes the machine performance and the network bandwidth, so that the synchronous processing mode can be adopted to solve the problems, and the system consumption and the network bandwidth can be saved. In practical application, the processing mode of the picture screening strategy can be selected based on system performance, network bandwidth and processing efficiency. Wherein system performance may be measured by at least one of latency, throughput, and scalability.

In an alternative aspect of the invention, the processing order is determined by:

The screening force is measured by the number of the screened pictures in a set time period, and the larger the number of the screened pictures is, the larger the screening force corresponding to the screening strategies is indicated. In practical application, an offline mode can be adopted to count the number of pictures screened by each screening strategy in a period of time, and the more the number of the screened pictures is, the earlier the processing sequence of the corresponding screening strategy is.

It can be understood that, in practical application, the screening strength of each screening policy is changed in real time, and then the processing sequence of each screening policy may be updated every set period of time by determining the processing sequence based on the screening capability.

In an alternative scheme of the invention, the method for determining the target cover from the candidate picture set based on the definition and/or the aesthetic degree of each picture in the candidate picture set comprises the following steps:

If the foregoing method of extracting the first picture set and the second picture set from the video to be processed is not adopted to determine the target cover, when determining the target cover based on the definition and/or the aesthetic degree of each picture in the candidate picture set, the following processing methods may be included:

first, based on the definition of each picture in the candidate picture set, the definition evaluation is performed on each picture to obtain a definition score (initial score) of each picture, and based on the initial score of each picture, a target cover is determined from the candidate picture set. One implementation scheme is as follows: and taking the picture with the highest definition score as the target cover whether the candidate picture set comprises the original cover picture or not. Or, based on the picture withholding strategy described above, withholding the initial score corresponding to each picture to obtain the final score of each picture, and based on the final score of each picture, determining the target cover from the candidate picture set.

Secondly, based on the attractiveness of each picture in the candidate picture set, carrying out attractiveness evaluation on each picture to obtain an attractiveness score (initial score) of each picture, and determining a target cover from the candidate picture set based on the initial score of each picture. The specific implementation manner may be the same as that of the first implementation manner, and will not be described herein.

Thirdly, based on the attractiveness and definition of each picture in the candidate picture set, respectively carrying out attractiveness evaluation and definition evaluation on each picture in the candidate picture set, wherein the obtained initial scores of each picture comprise an attractiveness score and a definition score, namely one picture corresponds to two scores, and based on the initial scores of each picture, determining a target cover from the candidate picture set. One implementation is: based on the aesthetic degree score and the definition score corresponding to each picture, determining the comprehensive score corresponding to each picture, and the same scheme as described above, whether the candidate picture set includes the original cover picture or not, the picture with the highest comprehensive score can be selected from the candidate picture set as the target cover directly based on the comprehensive score corresponding to each picture. Or, based on the picture withholding strategy described above, withholding the initial score corresponding to each picture to obtain the final score of each picture, and based on the final score of each picture, determining the target cover from the candidate picture set.

Wherein, based on the aesthetic degree score and the definition score corresponding to each picture, one realizable scheme for determining the comprehensive score corresponding to each picture is as follows: and determining the comprehensive score corresponding to each picture based on the weight corresponding to the aesthetic degree score and the weight corresponding to the definition and the aesthetic degree score and the definition score corresponding to each picture. Alternatively, the average score corresponding to the aesthetic score and the sharpness score is used as the composite score.

Each picture in the candidate picture set is scored twice, and the candidate picture set can comprise an original cover map or not; through the mode of determining the target cover, the picture can be screened aiming at two dimensions (attractive degree and definition) of one picture, so that the obtained target cover is more in line with the watching experience of a user.

In order to better understand the method provided by the embodiment of the present invention, the scheme of the embodiment of the present invention is further described in detail below with reference to fig. 2 and 3.

For a video release platform, the video to be processed can be the video to be released at the server side, and before the video to be processed is released, the target cover of the video to be processed can be determined based on the scheme in the server, so that the video after the target cover is determined has better video playing effect after being released, and the video can be more in line with the watching experience of a user. The following is a specific description of the solution of the invention based on the following examples:

Step 1: the whole flow of the cover diagram selecting strategy starts;

step 2: and extracting the video to be processed, wherein in the example, 20 frames of pictures are uniformly extracted from the video to be processed, 10 pictures are taken as a first picture set, 10 pictures are taken as a second picture set, no repeated pictures exist in the two picture sets, and the extracted pictures can be stored in a designated database for use in the subsequent steps. And, the original cover map of the video to be processed is obtained, and the extracted 20 pictures and the original cover map are used as an initial candidate picture set, namely an extraction map alternative pool shown in fig. 2.

Step 3: and carrying out definition scoring on 10 pictures in the first picture set and the original cover picture, namely 11 pictures in total, obtaining definition scoring of each picture, and storing the definition scoring of each picture in a designated database for use in subsequent steps.

The step 2 of extracting the first picture set and the step 3 of performing definition scoring on the first picture set and the original cover map correspond to the step 2 of extracting and scoring by a definition algorithm.

Step 4: and carrying out aesthetic degree scoring on 10 pictures in the second picture set and the original cover picture, namely 11 pictures in total, obtaining the aesthetic degree score of each picture, and storing the aesthetic degree score of each picture in a designated database for use in the subsequent steps.

The step 2 of extracting the second picture set and the step 4 of grading the second picture set and the original cover map in an aesthetic degree corresponds to the step 2 of extracting and grading the second picture set and the original cover map in an aesthetic degree algorithm.

Step 5: and (3) carrying out pornography detection on 21 pictures (20 extracted pictures and an original cover picture) in the step (2) based on a pornography screening strategy in the pornography module, determining the pornography score of each picture based on pornography information in each picture, and storing the pornography score of each picture in a specified database for use in the subsequent step. This step corresponds to the pornography model detection in fig. 2.

Step 6: and (3) performing temptation detection on 21 pictures (20 extracted pictures and an original cover picture) in the step (2) based on the temptation screening strategy in the temptation module, determining temptation scores of the pictures based on temptation information in each picture, and storing the temptation scores of the pictures in a specified database for use in subsequent steps. This step corresponds to the temptation model detection in fig. 2.

Step 7: and (3) detecting uncomfortable pictures of the 21 pictures (the extracted 20 pictures and the original cover picture) in the step (2) based on an uncomfortable picture screening strategy in the uncomfortable picture module, wherein the pictures containing uncomfortable content information are the uncomfortable pictures, and the pictures containing the uncomfortable content information such as nausea teeth, hair, acnes, snakes and the like are the uncomfortable pictures. The discomfort graph detection results (whether or not the discomfort graph is) of the pictures are stored in a specified database for use in the subsequent steps. This step corresponds to the discomfort graph model detection in fig. 2.

Step 8: and (3) detecting advertisement information of 21 pictures (20 extracted pictures and an original cover map) in the step (2) based on an advertisement screening strategy in the advertisement module, wherein the advertisement information is included in the pictures, namely the advertisement map. The advertisement information detection results (whether or not the advertisement image) of the respective pictures are stored in a specified database for use in the subsequent steps. This step corresponds to the advertising map model detection in fig. 2.

Step 9: and (3) carrying out mosaic information detection on 21 pictures (20 extracted pictures and an original cover picture) in the step (2) based on a mosaic screening strategy in a mosaic module, wherein the pictures comprise mosaic information, namely the pictures comprising the mosaic. The mosaic information detection results (whether or not the pictures contain mosaics) of the respective pictures are stored in a specified database for use in the subsequent steps. This step corresponds to the mosaic model detection in fig. 2.

Step 10: and (3) carrying out black-and-white detection on 21 pictures (the extracted 20 pictures and the original cover picture) in the step (2) based on a black-and-white strategy in the black-and-white module. The black and white edge detection result (whether or not the picture contains black and white edges) of each picture is stored in a designated database for use in the subsequent step. This step corresponds to the black and white edge model detection in fig. 2.

Step 11: and (3) carrying out solid color region detection on the 21 pictures (the extracted 20 pictures and the original cover map) in the step (2) based on a solid color strategy in the solid color module. The solid color detection results (whether or not the pictures contain solid color areas) of the respective pictures are stored in a specified database for use in the subsequent steps. This step corresponds to the pure chroma model detection in fig. 2.

In the steps 2 to 11, the processing results (detection results) of the pictures are stored in the designated database, so that the time consumption of the service can be effectively shortened, the utilization rate of the processing results can be improved, and the machine cost can be saved by using the caching strategy for the repeated calculation of the repeated pictures.

It should be noted that, the above-mentioned steps 2 to 11 are not limited to the above-mentioned one processing sequence, and each step may be performed in parallel or in other sequences, and may be configured based on actual requirements.

Step 12: as shown in fig. 2, the candidate pictures (the extracted 20 pictures and the original cover picture) are removed by the picture selection module according to the detection result of the above model, a candidate picture set is obtained, and the optimal cover picture (the target cover) is output from the candidate picture set based on the deduction strategy.

In this example, the screening principles of the individual screening policies may include, but are not limited to:

and if the cover map is detected to be the advertisement map, eliminating the advertisement map from the alternative map.

If the cover map is detected as the uncomfortable map, the cover map is removed from the alternative map.

And if the cover map detection pornography score exceeds a certain threshold value, eliminating the cover map detection pornography score from the alternative map.

And if the cover map detection temptation score exceeds a certain threshold value, eliminating from the alternative maps.

If the cover map detects a mosaic, or if the cover map detects a mosaic and the area occupation ratio of the mosaic exceeds a certain threshold, the mosaic is removed from the alternative map.

Step 13: if the candidate picture set does not comprise the original cover map, selecting a new cover map (target cover) from the candidate picture set, and if the candidate picture set comprises the original cover map, comparing and grading the original cover map with the optimal map selected from the candidate picture set, wherein the top grade is used as the target cover.

Alternatively, a specific selection strategy may be as follows:

polling and selecting pictures in the first picture set: the advertisement picture is not selected (screened out), the discomfort picture is not selected, the pornography exceeds the threshold value, the temptation exceeds the threshold value, the mosaic area ratio exceeds the threshold value, then the rest pictures are scored according to the definition in the step 3, and the picture with the highest score is selected as a first alternative picture (clearCover).

Polling and selecting pictures in the second picture set: the advertisement picture is not selected, the discomfort picture is not selected, the pornography degree exceeds the threshold value and the temptation degree exceeds the threshold value, the mosaic area ratio exceeds the threshold value and is not selected, then the rest pictures are scored according to the aesthetic degree in the step 4, and the picture with the highest score is selected as a second candidate picture (beautyyCover).

If the candidate picture set does not include the original cover map (original cover), outputting the clearCover as a formal definition cover map (first alternative picture), and outputting the beauty cover as a formal aesthetic cover map (second alternative picture); in order to better observe the display effect of the cover images recommended by the two algorithms of definition and attractiveness, the target cover images can be randomly selected from the first candidate images and the second candidate images according to the principle of random numbers.

If the candidate set of pictures includes an original cover map (origin cover), the sharpness scores of the origin cover and clearCover, and the aesthetics scores of the origin cover and beauty cover may be compared, respectively, and before comparing the scores, the three pictures may be subjected to a scoring process based on a picture scoring policy:

if the picture has black and white edges, the original score (definition score, aesthetic degree score and comprehensive score) of the picture is buckled for 3 points. If the solid-color image area ratio in the picture is too high, the original score is buckled by 3 scores.

And comparing the scores after the deduction, selecting one picture from the original cover and the clearCover as a first alternative picture, selecting one picture from the original cover and the beauty cover as a second alternative picture, and then selecting the target cover from the first alternative picture and the second book picture.

In this example, selecting the target cover map from two pictures (the first alternative picture and the second alternative picture) may be achieved by:

and determining the target cover from the first candidate picture and the second candidate picture based on a selection strategy corresponding to the current time. The selection policy may be determined according to the scheme described above, and will not be described herein.

After the target cover is determined, the video to be processed can be distributed into a video interaction platform, the PV value and/or the VV value of the video in a period of time are counted, and the playing effect of the video is determined based on the PV value and/or the VV value.

As an example, as a schematic diagram of the click rate (PV value) of the information stream (feeds referred to in the figure) shown in fig. 3, in which the horizontal axis is date, the vertical axis is click rate (PV value), the video to be processed is video a in this example, the thicker curve corresponds to the click rate change curve of the video a when a sharpness picture is the video cover of the video a, the thinner curve corresponds to the click rate change curve of the video a when an aesthetic picture is the video cover of the video a, the graph shows the click rate of the video a selected on the basis of sharpness on the 5 th day 2019 (20190805 in the figure) to the 12 th day 2019 (20190812 in the figure), the floating range is between 10.50% and 11.58%, and the click rate of the video a selected on the basis of the target cover of the aesthetic degree on the 5 th day 2019 (20190805 in the figure) to the 12 th day 2019 (20190812 in the figure), and the floating range is between 11.59% and 12.25%.

Based on the two click rates, it can be observed which impression of the target cover of the video a is better, which is determined based on sharpness and aesthetics.

Step 14: if the determined target cover includes a watermark or does not conform to the size of the target cover, the target cover may be subjected to a watermark removal process, and cut into a picture conforming to the size, where the step corresponds to the cover map in fig. 2 for watermark removal and cutting.

Step 15: and finally, manually checking the target cover determined in the step 14 to ensure that the target cover is more in line with the viewing experience of the user, pushing the target cover which is subjected to the manual checking to a user terminal (corresponding to the recommendation side in fig. 2), and displaying the target cover as the user.

The processing manners in the steps 5 to 9 are concurrent processing manners, that is, the modules are mutually independent and are in a decoupled state, and the processing manners in the steps 5 to 9 can also be synchronous processing manners, that is, the modules have a dependency relationship, and particularly, the schematic diagram of the synchronous processing manner shown in fig. 4 can be referred to.

In this example, the processing order of each module may be determined based on the number of the filtered pictures of each module history, and in this example, taking the number of pictures deleted due to uncomfortable diagrams > the number of pictures deleted due to mosaics > the number of pictures deleted due to excessive pornography > the number of pictures deleted due to excessive temptation > the number of pictures deleted due to advertisements as an example, the processing order of each filtering policy is as follows in order from first to last: a discomfort map screening strategy, a mosaic screening strategy, a pornography screening strategy, a temptation screening strategy and an advertisement screening strategy.

Based on the above screening policy, as shown in fig. 4, the 21 pictures in the step 2 may be screened by the uncomfortable picture screening policy, after a part of the pictures are filtered by the uncomfortable picture screening policy, the remaining pictures are screened by the mosaic screening policy, and so on, and finally the candidate picture set screened by the advertisement screening policy is obtained.

The subsequent steps are consistent with the processing manners of other steps, and are not described in detail herein.

According to the method, when the target cover of the video is determined, the video to be processed and the original cover image are screened based on at least two image screening strategies, so that the images can be screened from at least two dimensions to obtain candidate image sets, then the candidate image sets are screened through the definition and/or the attractiveness of each image, namely, the images are processed from another dimension, so that the finally obtained target cover of the video to be processed considers the influence of factors of multiple dimensions on the quality of the cover, and the determined target cover is higher in quality and accords with the watching experience of a user. And based on the video cover determination scheme, the manual auditing passing rate can be indirectly improved, and the human auditing cost is reduced.

Based on the same principle as the method shown in fig. 1, the embodiment of the present invention further provides a video cover determining apparatus 20, as shown in fig. 5, the video cover determining apparatus 20 may include a text video acquisition module 210, a picture screening module 220, and a cover determining module 230, wherein:

the video acquisition module 210 is configured to acquire a video to be processed and an original cover map of the video to be processed;

the picture screening module 220 is configured to screen a candidate picture set of the target cover from the video to be processed and the original cover map based on at least two picture screening policies;

the cover determining module 230 is configured to determine the target cover from the candidate picture set based on the sharpness and/or aesthetic degree of each picture in the candidate picture set.

Optionally, the picture screening module 220 is specifically configured to, when screening the candidate picture set of the target cover from the to-be-processed video and the original cover map based on at least two picture screening policies:

Optionally, if the candidate picture set includes the original cover map, the picture screening module 220 is specifically configured to, when determining the first candidate picture from the third picture set based on the sharpness of each picture in the third picture set:

the picture screening module 220 is specifically configured to, when determining the second candidate picture from the fourth picture set based on the aesthetic degree of each picture in the fourth picture set:

Optionally, the cover determining module 230 is specifically configured to, when determining the target cover from the candidate set of pictures based on the sharpness and/or aesthetic degree of each picture in the candidate set of pictures:

Optionally, the cover determining module 230 is specifically configured to, when determining the target cover from the candidate picture set based on the initial scores of the pictures:

Optionally, the picture screening policies include at least two of a pornography screening policy, a temptation screening policy, a discomfort map screening policy, an advertisement screening policy, or a mosaic screening policy.

or alternatively;

Optionally, the processing order is determined by:

Optionally, if the picture screening policy includes at least one of a pornography screening policy and an temptation screening policy, the picture screening module 220 is specifically configured to, when screening the candidate picture set of the target cover from the video to be processed and the original cover map based on at least two picture screening policies:

if the picture screening policy includes at least one of an uncomfortable picture screening policy, an advertisement screening policy and a mosaic screening policy, the picture screening module 220 is specifically configured to, when screening a candidate picture set of the target cover from the to-be-processed video and the original cover map based on at least two picture screening policies:

Since the video cover determining apparatus provided in the embodiment of the present invention is an apparatus capable of executing the video cover determining method in the embodiment of the present invention, a person skilled in the art can understand the specific implementation of the video cover determining apparatus in the embodiment of the present invention and various modifications thereof based on the video cover determining method provided in the embodiment of the present invention, so how the video cover determining apparatus implements the video cover determining method in the embodiment of the present invention will not be described in detail herein. The video cover determining device adopted by the video cover determining method in the embodiment of the invention belongs to the scope of protection required by the invention as long as the person skilled in the art implements the video cover determining device.

Based on the same principle as the video cover determining method and the video cover determining device provided by the embodiment of the invention, the embodiment of the invention also provides an electronic device, which can comprise a processor and a memory. The memory stores readable instructions that, when loaded and executed by the processor, implement the methods described in any of the embodiments of the present invention.

As an example, a schematic structural diagram of an electronic device 4000 to which the scheme of the embodiment of the present invention is applied is shown in fig. 6, and as shown in fig. 6, the electronic device 4000 may include a processor 4001 and a memory 4003. Wherein the processor 4001 is coupled to the memory 4003, such as via a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004. It should be noted that, in practical applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 is not limited to the embodiment of the present invention.

The processor 4001 may be a CPU (Central Processing Unit ), general purpose processor, DSP (Digital Signal Processor, data signal processor), ASIC (Application Specific Integrated Circuit ), FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic device, transistor logic device, hardware components, or any combination thereof. Which may implement or perform the various exemplary logic blocks, modules and circuits described in connection with this disclosure. The processor 4001 may also be a combination that implements computing functionality, e.g., comprising one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

Bus 4002 may include a path to transfer information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus or an EISA (Extended Industry Standard Architecture ) bus, or the like. The bus 4002 can be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 6, but not only one bus or one type of bus.

Memory 4003 may be, but is not limited to, ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, RAM (Random Access Memory ) or other type of dynamic storage device that can store information and instructions, EEPROM (Electrically Erasable Programmable Read Only Memory ), CD-ROM (Compact Disc Read Only Memory, compact disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer.

The memory 4003 is used for storing application program codes for executing the inventive arrangements, and is controlled to be executed by the processor 4001. The processor 4001 is configured to execute application code stored in the memory 4003 to implement the scheme shown in any of the method embodiments described above.

It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited in order and may be performed in other orders, unless explicitly stated herein. Moreover, at least some of the steps in the flowcharts of the figures may include a plurality of sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, the order of their execution not necessarily being sequential, but may be performed in turn or alternately with other steps or at least a portion of the other steps or stages.

The foregoing is only a partial embodiment of the present invention, and it should be noted that it will be apparent to those skilled in the art that modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations should and are intended to be comprehended within the scope of the present invention.

Claims

1. A method for determining a video cover, comprising:

extracting a plurality of pictures from the video to be processed, wherein one part of the pictures is used as a first picture set, and the other part of the pictures is used as a second picture set; screening the first picture set and the original cover map based on definition, screening the second picture set and the original cover map based on attractive appearance, and taking the screened pictures as first candidate picture sets;

screening a second candidate picture set of the target cover from the first candidate picture set based on at least two picture screening strategies, including: screening the first candidate picture set based on each of at least two picture screening strategies, wherein the pictures screened by each screening strategy are taken as second candidate picture sets of the target cover, and the pictures screened by each screening strategy are mutually independent; or, acquiring a processing sequence of each screening strategy in the at least two image screening strategies, and sequentially screening the first candidate image set according to the processing sequence to obtain a second candidate image set, wherein the current strategy is used for screening the images screened by the previous strategy;

Determining the target cover from the second candidate picture set based on the definition and/or the aesthetic degree of each picture in the second candidate picture set, including: scoring each picture in the second candidate picture set based on the definition and/or the attractiveness of the picture, performing deduction based on a preset picture deduction strategy, and determining the target cover from the second candidate picture set based on a deduction result; the picture deduction strategy is related to the influence degree of black and white edges and/or solid-color image areas in the picture on the video playing effect;

the method further comprises the steps of: and storing the processing results of the strategies in a preset database so as to directly acquire corresponding processing results from the database when the same picture is processed.

2. The method of claim 1, wherein the determining the target cover from the second candidate picture set based on the sharpness and/or aesthetics of each picture in the second candidate picture set comprises:

determining a first alternative picture from a third picture set based on the definition of each picture in the third picture set;

determining a second alternative picture from a fourth picture set based on the aesthetic degree of each picture in the fourth picture set;

Determining the target cover from the first candidate picture and/or the second candidate picture;

if the second candidate picture set includes the original cover map, the third picture set includes the picture in the first picture set and the original cover map in the second candidate picture set, and the fourth picture set includes the picture in the second picture set and the original cover map in the second candidate picture set;

if the second candidate picture set does not include the original cover map, the third picture set is a picture in the first picture set in the candidate picture set, and the fourth picture set is a picture in the second picture set in the second candidate picture set.

3. The method of claim 2, wherein if the second candidate picture set includes the original cover map, the determining the first candidate picture from the third picture set based on the sharpness of each picture in the third picture set comprises:

The determining, based on the aesthetic degree of each picture in the fourth picture set, a second candidate picture from the fourth picture set includes:

and screening a second candidate picture from the second candidate picture and the original cover picture based on the attractive degree of the second candidate picture, the attractive degree of the original cover picture and a preset picture deduction strategy.

4. The method of claim 1, wherein the determining the target cover from the second candidate picture set based on the sharpness and/or aesthetics of each picture in the second candidate picture set comprises:

based on the definition and/or the aesthetic degree of each picture in the second candidate picture set, respectively evaluating each picture in the second candidate picture set to obtain an initial score of each picture;

and determining the target cover from the second candidate picture set based on the initial scores of the pictures.

5. The method of claim 4, wherein the determining the target cover from the second candidate set of pictures based on the initial scores of the pictures comprises:

and determining the target cover from the second candidate picture set based on the final scores of the pictures.

6. The method of any one of claims 1 to 5, wherein the picture screening policies include at least two of a pornography screening policy, an temptation screening policy, a discomfort map screening policy, an advertising screening policy, or a mosaic screening policy.

7. The method of claim 1, wherein the processing order is determined by:

8. The method of claim 6, wherein if the picture screening policy includes at least one of the pornography screening policy and the temptation screening policy, the screening the second candidate picture set of the target cover from the first candidate picture set based on at least two picture screening policies comprises:

Determining the yellow level of each picture in the first candidate picture set;

deleting pictures with yellow grades exceeding a set grade from the first candidate picture set, and taking the remaining pictures after deletion as the second candidate picture set;

if the picture screening policy includes at least one of the discomfort map screening policy, the advertisement screening policy and the mosaic screening policy, the screening the second candidate picture set of the target cover from the first candidate picture set based on at least two picture screening policies includes:

detecting whether the first candidate picture set comprises setting information, wherein the setting information comprises at least one of uncomfortable content information, advertisement information and mosaic information;

deleting the pictures containing the setting information from the first candidate picture set, and taking the remaining pictures after deletion as the second candidate picture set.

9. A video cover determining apparatus, comprising:

the video acquisition module is used for acquiring a video to be processed and an original cover map of the video to be processed;

the picture screening module is used for extracting a plurality of pictures from the video to be processed, wherein one part of the pictures is used as a first picture set, and the other part of the pictures is used as a second picture set; screening the first picture set and the original cover map based on definition, screening the second picture set and the original cover map based on attractive appearance, and taking the screened pictures as first candidate picture sets; screening a second candidate picture set of the target cover from the first candidate picture set based on at least two picture screening strategies, including: screening the first candidate picture set based on each of at least two picture screening strategies, wherein the pictures screened by each screening strategy are taken as second candidate picture sets of the target cover, and the pictures screened by each screening strategy are mutually independent; or, acquiring a processing sequence of each screening strategy in the at least two image screening strategies, and sequentially screening the first candidate image set according to the processing sequence to obtain a second candidate image set, wherein the current strategy is used for screening the images screened by the previous strategy;

The cover determining module is configured to determine, based on the sharpness and/or aesthetic degree of each picture in the second candidate picture set, the target cover from the second candidate picture set, and includes: scoring each picture in the second candidate picture set based on the definition and/or the attractiveness of the picture, performing deduction based on a preset picture deduction strategy, and determining the target cover from the second candidate picture set based on a deduction result; the picture deduction strategy is related to the influence degree of black and white edges and/or solid-color image areas in the picture on the video playing effect;

the device also comprises a module for storing the processing results of the strategies in a preset database so as to directly acquire corresponding processing results from the database when the same picture is processed.