WO2017177559A1 - 一种图像管理方法和装置 - Google Patents

一种图像管理方法和装置 Download PDF

Info

Publication number
WO2017177559A1
WO2017177559A1 PCT/CN2016/088611 CN2016088611W WO2017177559A1 WO 2017177559 A1 WO2017177559 A1 WO 2017177559A1 CN 2016088611 W CN2016088611 W CN 2016088611W WO 2017177559 A1 WO2017177559 A1 WO 2017177559A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
repeated
images
difference
repeated image
Prior art date
Application number
PCT/CN2016/088611
Other languages
English (en)
French (fr)
Inventor
陈军
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2017177559A1 publication Critical patent/WO2017177559A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10—File systems; File servers
    • G06F16/17—Details of further file system functions
    • G06F16/1737—Details of further file system functions for reducing power consumption or coping with limited storage space, e.g. in mobile devices
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10—File systems; File servers
    • G06F16/16—File or folder operations, e.g. details of user interfaces specifically adapted to file systems
    • G06F16/162—Delete operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10—File systems; File servers
    • G06F16/17—Details of further file system functions
    • G06F16/174—Redundancy elimination performed by the file system
    • G06F16/1748—De-duplication implemented within the file system, e.g. based on file segments
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules

Definitions

  • This application relates to, but is not limited to, the field of mobile communication technology.
  • This paper provides an image management method and device to solve the problem of managing the image of the mobile terminal, thereby effectively processing redundant photos and saving space.
  • An image management method comprising:
  • a duplicate image to be deleted is determined based on the difference weight of each repeated image in the repeated image.
  • the identifying the repeated images in the multiple images includes:
  • the determining, according to the difference weight of each repeated image in the repeated image, determining the repeated image to be deleted includes:
  • the repeated images with the smallest difference weights in the repeated images are retained, and the remaining repeated images are deleted.
  • the method further includes:
  • the method further includes:
  • determining, when the repeated image is a difference image, a difference weight of each repeated image in the repeated image including:
  • a difference weight of each repeated image in the repeated image is determined according to the difference image category and the corresponding difference weight.
  • An image management device comprising:
  • the obtaining module is set to: acquire multiple images created within a predetermined time period;
  • An identification module configured to: identify a repeated image in the plurality of images acquired by the acquiring module;
  • a classification module configured to: determine, according to the preset difference image category, whether the repeated image recognized by the recognition module is a poor image
  • An operation module configured to: when the classification module determines that the repeated image is a poor image, Determining a difference weight of each repeated image in the repeated image;
  • And deleting the module configured to: determine, according to the operation module, a difference weight of each repeated image in the repeated image, and determine a repeated image to be deleted.
  • the identifying module includes:
  • Determining a comparison unit configured to: determine a mean square error of pixels between the plurality of images created in the predetermined time period, and a mean square error of pixels between the plurality of images and a mean square error threshold of a preset pixel Compare;
  • a determining unit configured to: when the comparison result of the determining comparing unit is that a mean square error of a pixel between the plurality of images is smaller than a mean square error threshold of the preset pixel, determining that the multiple images are repeated images .
  • the deleting module includes:
  • a first deleting unit configured to: when the computing module determines that the difference weights of each repeated image in the repeated image are the same, retain one of the repeated images, and delete the remaining repeated images;
  • a second deleting unit configured to: when the operation module determines that the difference weight of each repeated image in the repeated image is different, retain a repeated image with the smallest difference weight in the repeated image, and delete the remaining repeated image.
  • the deleting module is further configured to: when the computing module determines that the repeated image is not a bad image, retain one of the repeated images, and delete the remaining repeated images.
  • the device further includes:
  • a weight setting module configured to: before the operation module determines a difference weight of each repeated image in the repeated image, preset a difference weight corresponding to the difference image category;
  • the operation module is configured to: when the classification module determines that the repeated image is a difference image, determine a difference weight of the repeated image, including:
  • the classification module determines that the repeated image is a difference image, determining the repeated image according to the difference image category and a difference weight corresponding to the difference image category set by the weight setting module. The difference weight of each repeated image.
  • An image management method and apparatus provided by an embodiment of the present invention are created by acquiring a predetermined time period a plurality of images, identifying a repeated image in the plurality of images, and determining whether the repeated image is a poor image according to a preset difference image type, thereby determining each of the plurality of images when the complex image is determined to be a poor image
  • the difference weight of the image is repeated, and the repeated image to be deleted is determined according to the difference weight of each repeated image.
  • the embodiment of the present invention solves the problem that the mobile terminal stores a large number of repeated images after taking a picture, and implements automatic deletion to remove the repeated image. It is convenient for user image management and release of mobile terminal memory space.
  • FIG. 1 is a flowchart of an image management method according to an embodiment of the present invention
  • FIG. 2 is a flowchart of another image management method according to an embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram of an image management apparatus according to an embodiment of the present disclosure.
  • FIG. 4 is a schematic structural diagram of an identification module in an image management apparatus according to an embodiment of the present invention.
  • FIG. 5 is a schematic structural diagram of another image management apparatus according to an embodiment of the present disclosure.
  • FIG. 6 is a schematic structural diagram of a deletion module in an image management apparatus according to an embodiment of the present invention.
  • FIG. 1 is a flowchart of an image management method according to an embodiment of the present invention. As shown in FIG. 1 , the image management method provided in this embodiment may include the following steps, namely, S110 to S140:
  • S110 acquiring a plurality of images created within a predetermined time period, and identifying repeated images in the plurality of images;
  • S120 Determine, according to the preset difference image category, whether the repeated image is a poor image
  • S140 Determine a repeated image to be deleted according to the difference weight of each repeated image in the repeated image.
  • a large amount of redundant data in a digital photo is a photo that is repeatedly taken by the user in order to obtain a better shooting effect in the same shooting scene.
  • the shooting time is close, therefore, the process of removing the duplicate photos can be performed after the collection time is concentrated on a certain continuous time period, and the continuous time period can be determined according to the user's selection, and can be set to 10 In seconds, it can also be set to 1 minute.
  • the image management method provided in this embodiment identifies a plurality of images created in a predetermined time period, identifies a repeated image in the plurality of images, and determines whether the repeated image is a poor image according to a preset difference image category. Therefore, when it is determined that the complex image is a difference image, the difference weight of each repeated image is determined, and the repeated image to be deleted is determined according to the difference weight of each repeated image; this embodiment solves the problem that the mobile terminal takes a picture The problem of storing a large number of repeated images realizes automatic deletion to remove duplicate images, facilitating image management of the user and releasing the memory space of the mobile terminal.
  • FIG. 2 is a flowchart of another image management method according to an embodiment of the present invention.
  • the implementation manner of identifying the repeated images in the multiple images in the embodiment, that is, S110 may include the following steps, namely, S111 to S112:
  • S111 Determine a mean square error of pixels between the plurality of images created in the predetermined time period, and compare a mean square error of the pixels between the multiple images with a mean square error threshold of the preset pixels.
  • the differential image category preset in S120 includes, for example, at least one of the following: ghosting, underexposure, portrait not in a preset position, closed eyes, etc., when repeated
  • the repeated image may be determined as a difference image, and otherwise, the repeated image may be determined not to be a difference image but a qualified image.
  • the image data of the photo can be input into a specific difference classifier for classification.
  • the classifier may be a classifier that has been trained in advance by a neural network algorithm, for example, may be large
  • the ghosting sample input classifier is trained to obtain the parameters of the classifier and configured into the storage device.
  • the storage device calculates the image to be discriminated by using these parameters and then determines whether it is the image of the category.
  • Convolutional neural networks can be used when selecting neural network models. They are suitable for identifying two-dimensional graphics with displacement, scaling and other forms of distortion invariance. Without the need to manually extract image features, it is possible to implicitly learn image features from training data and classify them. .
  • a specific part of the image to be discriminated may also be selected according to the category of the classifier to participate in the discrimination to improve the calculation speed. For example, if it is judged whether the image is a blinking photo, it is only necessary to select the data calculation of the face part, and the face positioning can be realized by the related art scheme, which can greatly reduce the calculation amount.
  • the duplicate photograph directly uses the same discriminating image region for classification and discrimination.
  • the embodiment of the present invention utilizes a neural network to train a large number of various badly-performing samples in advance, so that the smart device can learn these "bad" features and obtain different types of classifiers, thereby discriminating the user's image database and deleting redundant images. Photographs with poor shooting results automatically complete photo cleaning. Specific classifiers can separate "bad" photos according to different categories, such as determining which are ghosts and which are blinking. Optionally, one or more classifiers of different difference image categories may be used to determine whether the image has one or more different differential features, and the user may selectively classify the categories of interest to the user. It can be used alone or in combination with a poor image classifier.
  • the embodiment may further include:
  • the implementation manner of determining the repeated image to be deleted in the embodiment, that is, S130 may include: determining, according to the difference image category and the corresponding difference weight, the difference weight of each repeated image in the repeated image.
  • the difference weight of the repeated image is the sum of the weights of the difference types to which the repeated image conforms.
  • Each of the difference image categories may be assigned a difference weight, which may be preset by the user according to the tolerance of the subjective experience of the individual, for example: ghost image 30%, underexposure 10%, portrait not in the preset position 20 %, closed eyes 30%, and the like.
  • the weights corresponding to the two difference image types of ghost image and closed eye can be accumulated, that is, the image has more difference features.
  • the greater the difference weight the subjective visual effect
  • the simpler evaluation rule can be adopted in the embodiment of the present invention, and other well-known algorithms can also be used to perform accurate image difference weight calculation.
  • the implementation manner of determining the repeated image to be deleted in the embodiment, that is, the S140 may include the following steps, that is, S141 to S142:
  • the image management method provided by the embodiment of the present invention may further include: when the repeated image is not a bad image, only one of the repeated images is retained, and the remaining repeated images are deleted; for example, when all the repeated images do not have the bad image category; , then judged as a qualified image, at this time directly retain one, and the rest can be deleted.
  • FIG. 3 is a schematic structural diagram of an image management apparatus according to an embodiment of the present invention. As shown in FIG. 3, the image management apparatus provided in this embodiment includes:
  • the obtaining module 10 is configured to: acquire a plurality of images created within a predetermined time period;
  • the identification module 20 is configured to: identify a repeated image in the plurality of images acquired by the acquisition module 10;
  • the classification module 30 is configured to: determine, according to the preset difference image category, whether the repeated image recognized by the recognition module 20 is a poor image;
  • the operation module 40 is configured to: when the classification module 30 determines that the repeated image is a difference image, determine a difference weight of each repeated image in the repeated image;
  • the deleting module 50 is configured to: according to the computing module 40, determine a difference weight of each repeated image in the repeated image, and determine a repeated image to be deleted.
  • the preset difference image category includes, for example, at least one of the following: ghosting, underexposure, overexposure, portrait not in a preset position, closed eyes, and the like, when the repeated image exists
  • the repeated image may be determined as a difference image, otherwise, it is determined that the repeated image is not a difference image but a qualified image.
  • FIG. 4 is a schematic structural diagram of an identification module in an image management apparatus according to an embodiment of the present invention.
  • the identification module 20 in this embodiment may include:
  • the determining comparison unit 201 is configured to: determine a mean square error of pixels between the plurality of images created in the predetermined time period, and compare a mean square error of the pixels between the plurality of images with a mean square error threshold of the preset pixels;
  • the determining unit 202 is configured to determine that the plurality of images are repeated images when it is determined that the comparison result of the comparing unit 201 is that the mean square error of the pixels between the plurality of images is smaller than the mean square error threshold of the preset pixels.
  • the image data of the photo can be input into a specific difference classifier for classification.
  • the classifier may be a classifier that has been trained in advance by a neural network algorithm. For example, a sample of a large number of ghosts may be input to a classifier for training, and parameters of the classifier are obtained and configured into a storage device. The storage device calculates the image to be discriminated by using these parameters and then determines whether it is the image of the category.
  • Convolutional neural networks can be used when selecting neural network models. They are suitable for identifying two-dimensional graphics with displacement, scaling and other forms of distortion invariance. Without the need to manually extract image features, it is possible to implicitly learn image features from training data and classify them. .
  • a specific part of the image to be discriminated may also be selected according to the category of the classifier to participate in the discrimination to improve the calculation speed. For example, if it is judged whether the image is a blinking photo, it is only necessary to select the data calculation of the face part, and the face positioning can be realized by the related art scheme, which can greatly reduce the calculation amount.
  • the duplicate photograph directly uses the same discriminating image region for classification and discrimination.
  • the embodiment of the present invention utilizes a neural network to train a large number of various badly-performing samples in advance, so that the smart device can learn these "bad" features and obtain different types of classifiers, thereby discriminating the user's image database and deleting redundant images. Photographs with poor shooting results automatically complete photo cleaning. Specific classifiers can separate "bad" photos according to different categories, such as determining which are ghosts and which are blinking. Alternatively, one or more difference image classifiers may be employed, that is, one classifier only judges whether a photo effect is good or bad. Adopt more The classifiers can be classified by the user to selectively classify them, and can be used alone or in combination.
  • the cleaning may be completed without other types of determination. If the image is repeated, only one of the images is judged as pass by all the classifiers, and the remaining images in the repeated image can be directly cleaned without judgment.
  • FIG. 5 is a schematic structural diagram of another image management apparatus according to an embodiment of the present invention. Based on the structure of the device shown in FIG. 3, the image management device provided in this embodiment further includes:
  • the weight setting module 60 is configured to: before the computing module 40 determines the difference weight of each repeated image in the repeated image, preset the difference weight corresponding to the difference image category; according to the difference image category and the corresponding difference The weight determines the difference weight of each repeated image in the repeated image.
  • the operation module 40 is configured to determine, when the repeated image is a difference image, the difference weight of each repeated image in the repeated image is: when the classification module 30 determines that the repeated image is a difference image At the time, the difference weight of each repeated image in the repeated image is determined according to the difference weight corresponding to the difference image category set by the difference image category and weight setting module 60.
  • the difference weight of the repeated image is the sum of the weights of the difference types to which the repeated image conforms.
  • Each of the difference image categories may be assigned a difference weight, which may be preset by the user according to the tolerance of the subjective experience of the individual, for example: ghost image 30%, underexposure 10%, portrait not in the preset position 20 %, closed eyes 30%, and the like.
  • the weights corresponding to the two difference image types of ghost image and closed eye can be accumulated, that is, the image has more difference features.
  • the larger the difference weight the worse the subjective visual effect; the simpler evaluation rule can be adopted in the embodiment of the present invention, and other well-known algorithms can also be used to perform accurate image difference weight calculation.
  • FIG. 6 is a schematic structural diagram of a deletion module in an image management apparatus according to an embodiment of the present disclosure.
  • the deletion module 50 in this embodiment may include:
  • the first deleting unit 501 is configured to: when the computing module 40 determines that the difference weights of each repeated image in the repeated image are the same, retain one of the repeated images, and delete the remaining repeated images;
  • the second deleting unit 502 is configured to: when the computing module 40 determines each repeated image in the repeated image When the difference weights of the images are different, the repeated images with the smallest difference weights in the repeated images are retained, and the remaining repeated images are deleted.
  • the deleting module 50 is further configured to: when the computing module 40 determines that the repeated image is not a bad image, retain one of the repeated images, and delete the remaining repeated images. For example, when the difference weights of each repeated image in the repeated image are the same, one of the repeated images is retained, and the remaining repeated images are deleted; when the difference weights of each repeated image in the repeated image are different, the repeated images are retained. The repeated image with the smallest difference weight removes the remaining duplicate image. When all of the duplicate images have the same type of difference image, for example, all of them are closed eyes, one is directly retained, and the remaining images are deleted.
  • the difference weights of each repeated image are different, for example, one is closed eyes and the other is closed eyes and underexposed, the quality of the second sheet is obviously worse than the first one, and the best quality is retained. , delete the rest of the photos. If the repeated image is not a poor image, only one of the repeated images is retained, and the remaining repeated images are deleted; specifically, when all the repeated images do not have the difference image category, the image is a qualified image, and one is directly retained at this time. The rest can be deleted.
  • all or part of the steps of the above embodiments may also be implemented by using an integrated circuit. These steps may be separately fabricated into individual integrated circuit modules, or multiple modules or steps may be fabricated into a single integrated circuit module. achieve.
  • the devices/function modules/functional units in the above embodiments may be implemented by a general-purpose computing device, which may be centralized on a single computing device or distributed over a network of multiple computing devices.
  • the device/function module/functional unit in the above embodiment When the device/function module/functional unit in the above embodiment is implemented in the form of a software function module and sold or used as a stand-alone product, it can be stored in a computer readable storage medium.
  • the above mentioned computer readable storage medium may be a read only memory, a magnetic disk or an optical disk or the like.
  • the embodiment of the present invention identifies a plurality of images created in a predetermined time period, identifies a repeated image in the plurality of images, and determines whether the repeated image is a poor image according to a preset difference image type, thereby determining When the complex image is a difference image, the difference weight of each repeated image is determined, and the repeated image to be deleted is determined according to the difference weight of each repeated image.
  • the embodiment of the present invention solves the problem that the mobile terminal stores a large number of repetitions after taking the image. Image problems, automatic deletion to remove duplicate images, convenient user image management and release of mobile terminal memory space.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Human Computer Interaction (AREA)
  • Image Analysis (AREA)

Abstract

一种图像管理方法和装置，其中，该方法包括：获取预定时间段内创建的多张图像，并识别所述多张图像中的重复图像；根据预设的差质图像类别，判断所述重复图像是否为差质图像；当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重；根据所述重复图像中每张重复图像的差质权重，确定待删除的重复图像。

Description

一种图像管理方法和装置 技术领域
本申请涉及但不限于移动通信技术领域。
背景技术
手机、数码相机等设备满足了用户随时随地移动拍摄的需求，但也极大的占用了大量的存储空间。由于大部分用户不会经常随时整理照片，这里面往往有大量的“效果不好”的照片，例如，存储有大量的同一场景下重复拍摄的照片，有些照片具有重影、人物照中出现眨眼等问题。这些照片无需长期保存，但是依靠用户主动整理、清理又费时费力，所以如果智能设备能帮助用户智能清理照片或者提供智能辅助分类，比如删除或标识出重复的、重影的、眨眼的、非正常曝光等“效果不好”的冗余照片，将极大节省用户整理照片的时间和精力，也能极大节省设备的存储空间资源。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本文提供一种图像管理方法和装置，以解决移动终端图像的管理问题，从而能够有效的处理冗余照片，节约空间。
一种图像管理方法，包括：
获取预定时间段内创建的多张图像，并识别所述多张图像中的重复图像；
根据预设的差质图像类别，判断所述重复图像是否为差质图像；
当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重；
根据所述重复图像中每张重复图像的差质权重，确定待删除的重复图像。
可选地，所述识别所述多张图像中的重复图像，包括：
确定所述预定时间段内创建的所述多张图像之间像素的均方差，将所述多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
当所述多张图像之间的像素的均方差小于所述预设的像素的均方差阈值时，判定所述多张图像为重复图像。
可选地，所述根据所述重复图像中每张重复图像的差质权重，确定待删除的重复图像，包括：
当所述重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
当所述重复图像中每张重复图像的差质权重不同时，保留所述重复图像中差质权重最小的重复图像，删除剩余重复图像。
可选地，所述方法还包括：
当所述重复图像不是差质图像时，保留其中一张重复图像，删除剩余重复图像；
可选地，所述当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重之前，所述方法还包括：
预设所述差质图像类别对应的差质权重；
所述当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重，包括：
根据所述差质图像类别和对应的差质权重，确定所述重复图像中每张重复图像的差质权重。
一种图像管理装置，包括：
获取模块，设置为：获取预定时间段内创建的多张图像；
识别模块，设置为：识别所述获取模块获取的所述多张图像中的重复图像；
分类模块，设置为：根据预设的差质图像类别，判断所述识别模块识别出的所述重复图像是否为差质图像；
运算模块，设置为：当所述分类模块判断出所述重复图像为差质图像时， 确定所述重复图像中每张重复图像的差质权重；
删除模块，设置为：根据所述运算模块确定出所述重复图像中每张重复图像的差质权重，确定待删除的重复图像。
可选地，所述识别模块包括：
确定比较单元，设置为：确定所述预定时间段内创建的所述多张图像之间像素的均方差，将所述多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
判定单元，设置为：当所述确定比较单元的比较结果为所述多张图像之间的像素的均方差小于所述预设的像素的均方差阈值时，判定所述多张图像为重复图像。
可选地，所述删除模块包括：
第一删除单元，设置为：当所述运算模块确定出所述重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
第二删除单元，设置为：当所述运算模块确定出所述重复图像中每张重复图像的差质权重不同时，保留所述重复图像中差质权重最小的重复图像，删除剩余重复图像。
可选地，所述删除模块，还设置为：当所述运算模块确定出所述重复图像不是差质图像时，保留其中一张重复图像，删除剩余重复图像。
可选地，所述装置还包括：
权重设置模块，设置为：在所述运算模块确定所述重复图像中每张重复图像的差质权重之前，预设所述差质图像类别对应的差质权重；
所述运算模块设置为当所述分类模块判断出所述重复图像为差质图像时，确定所述重复图像的差质权重，包括：
当所述分类模块判断出所述重复图像为差质图像时，根据所述差质图像类别和所述权重设置模块设置的与所述差质图像类别对应的差质权重，确定所述重复图像中每张重复图像的差质权重。
本发明实施例提供的图像管理方法和装置，通过获取预定时间段内创建 的多张图像，识别该多张图像中的重复图像，并根据预设的差质图像类别，判断上述重复图像是否为差质图像，从而在判断出复图像为差质图像时，确定每张重复图像的差质权重，并根据每张重复图像的差质权重，确定待删除的重复图像；本发明实施例解决了移动终端拍照后存储有大量重复图像的问题，实现自动删除以去除重复图像，方便用户的图像管理和移动终端内存空间的释放。
在阅读并理解了附图和详细描述后，可以明白其他方面。
附图概述
图1为本发明实施例提供的一种图像管理方法的流程图；
图2为本发明实施例提供的另一种图像管理方法的流程图
图3为本发明实施例提供的一种图像管理装置的结构示意图；
图4为本发明实施例提供的图像管理装置中一种识别模块的结构示意图；
图5为本发明实施例提供的另一种图像管理装置的结构示意图；
图6为本发明实施例提供的图像管理装置中一种删除模块的结构示意图。
本发明的实施方式
下文中将结合附图对本发明的实施方式进行详细说明。需要说明的是，在不冲突的情况下，本文中的实施例及实施例中的特征可以相互任意组合。
在附图的流程图示出的步骤可以在诸根据一组计算机可执行指令的计算机系统中执行。并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
图1为本发明实施例提供的一种图像管理方法的流程图。如图1所示，本实施例提供的图像管理方法可以包括如下步骤，即S110～S140：
S110：获取预定时间段内创建的多张图像，并识别该多张图像中的重复图像；
S120：根据预设的差质图像类别，判断重复图像是否为差质图像；
S130：当重复图像为差质图像时，确定重复图像中每张重复图像的差质权重；
S140：根据重复图像中每张重复图像的差质权重，确定待删除的重复图像。
通常地，数码照片中的大量冗余数据是在同一拍摄场景下，用户为了获得较好拍摄效果而重复拍摄的照片。显然这些重复的照片，拍摄时间是接近的，因此，可以按照创建时间在某一个连续时段上集中后进行去除重复照片的处理，这个连续的时间段可以根据用户的选择进行确定，可以设置为10秒钟，也可以设置为1分钟。
本实施例提供的图像管理方法，通过获取预定时间段内创建的多张图像，识别该多张图像中的重复图像，并根据预设的差质图像类别，判断上述重复图像是否为差质图像，从而在判断出复图像为差质图像时，确定每张重复图像的差质权重，并根据每张重复图像的差质权重，确定待删除的重复图像；本实施例解决了移动终端拍照后存储有大量重复图像的问题，实现了自动删除以去除重复图像，方便用户的图像管理和移动终端内存空间的释放。
可选地，图2为本发明实施例提供的另一种图像管理方法的流程图。在上述图1所示实施例的基础上，本实施例中识别多张图像中的重复图像的实现方式，即S110可以包括如下步骤，即S111～S112：
S111，确定预定时间段内创建的多张图像之间像素的均方差，将多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
S112，当多张图像之间的像素的均方差小于预设的像素的均方差阈值时，判定该多张图像为重复图像。
可选地，在本发明的一个实施例中，在S120中预设的差质图像类别例如包括如下至少之一：重影、曝光不足、人像未在预设位置、闭眼等等，当重复图像存在上述一种或者多种差质图像类别时，则可以将重复图像判定为差质图像，否则，判定重复图像不是差质图像而是合格图像。
在实际应用中，可以将相片的图像数据输入特定差质分类器进行分类。分类器可以是预先通过神经网络算法训练过的分类判别器，例如，可以将大 量重影的样张输入分类器进行训练，得到分类器的参数并配置到存储设备中。存储设备对待判别图像采用这些参数进行计算后来判定是否是该类别图像。选择神经网络模型时可以采用卷积神经网络，其适合用来识别位移、缩放及其他形式扭曲不变性的二维图形，无需人工提取图像特征，能够隐式地从训练数据中学习图像特征并分类。
实际应用中，还可根据分类器的类别，来选择待判别图像的特定部分来参与判别，以提高计算速度。例如判断图像是否是眯眼照片，则只需选择人脸部分的数据计算即可，而人脸定位采用相关技术的方案即可实现，这样可以大幅减少计算量。特别是对重复照片，如果前面已有照片确定了需要判别的图像区域，则该重复照片则直接采用相同的判别图像区域进行分类判别。本发明实施例就是利用神经网络事先训练海量的各种效果不好的样张，使得智能设备能学习这些“坏”特征，获得的不同类别的分类器，从而去判别用户的图像数据库，删除重复多余的、拍摄效果差的相片，自动完成相片清理工作。特定分类器是可以将“效果不好”的相片按照不同类别进行分开，比如判断哪些是重影的，哪些是人物眯眼的等等。可选地，可以采用一个或多个不同差质图像类别的分类器，来判别图像是否具有某种或多种不同的差质特征，可以由用户选择性地对其关注的类别进行分类处理，即可单独使用也可组合使用差质图像分类器。
可选地，在上述实施例的基础上，本实施例在S130之前还可以包括：
S121，预设差质图像类别对应的差质权重。
相应地，本实施例中确定待删除的重复图像的实现方式，即S130可以包括：根据差质图像类别和对应的差质权重，确定重复图像中每张重复图像的差质权重。
在本实施例中，重复图像的差质权重即本重复图像所符合的差质类型的权重之和。可以将每一差质图像类别对应一个差质权重，这个权重评分可以由用户根据个人主观感受的容忍度进行预设，例如：重影30％、曝光不足10％、人像未在预设位置20％、闭眼30％等诸如此类。当一张图像中具有重影和闭眼两个差质图像类型时，可以将重影和闭眼两个差质图像类型对应的权重进行累加，也就是说图像具有的差质特征越多，其差质权重越大，主观视觉效果 越差；本发明实施例可以采用的是较为简单的评判规则，也可以采用其他公知的算法进行精确的图像差质权重计算。
可选地，在上述实施例的基础上，本实施例中确定待删除的重复图像的实现方式，即S140可以包括如下步骤，即S141～S142：
S141，当重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
S142，当重复图像中每张重复图像的差质权重不同时，保留重复图像中差质权重最小的重复图像，删除剩余重复图像。
在实际应用中，当所有的重复图像的差质图像类型相同，例如，全部为闭眼，则直接保留一张，删除剩余的图像。当每张重复图像的差质权重不同时，例如，一张为闭眼，另外一张为闭眼和曝光不足，则第二张的质量显然比第一张更差，则保留质量相对最好的，删除其余照片。
本发明实施例提供的图像管理方法还可以包括：当重复图像不是差质图像，仅保留其中一张重复图像，删除剩余重复图像；举例来说，当所有重复图像都不具有差质图像类别时，则判定为合格图像，此时直接保留一张，其余删除即可。
图3为本发明实施例提供的一种图像管理装置的结构示意图，如图3所示，本实施例提供的图像管理装置包括：
获取模块10，设置为：获取预定时间段内创建的多张图像；
识别模块20，设置为：识别获取模块10获取的多张图像中的重复图像；
分类模块30，设置为：根据预设的差质图像类别，判断识别模块20识别出的重复图像是否为差质图像；
运算模块40，设置为：当分类模块30判断出重复图像为差质图像时，确定重复图像中每张重复图像的差质权重；
删除模块50，设置为：根据运算模块40确定出重复图像中每张重复图像的差质权重，确定待删除的重复图像。
可选地，本发明实施例中，预设的差质图像类别例如包括如下至少之一：重影、曝光不足、过曝、人像未在预设位置、闭眼等等，当重复图像存在上述一种或者多种差质图像类别时，则可以将重复图像判定为差质图像，否则，判定重复图像不是差质图像而是合格图像。
可选地，图4为本发明实施例提供的图像管理装置中一种识别模块的结构示意图，如图4所示，本实施例中的识别模块20可以包括：
确定比较单元201，设置为：确定预定时间段内创建的多张图像之间像素的均方差，将多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
判定单元202，设置为：当确定比较单元201的比较结果为多张图像之间的像素的均方差小于预设的像素的均方差阈值时，判定多张图像为重复图像。
在实际应用中，可以将相片的图像数据输入特定差质分类器进行分类。分类器可以是预先通过神经网络算法训练过的分类判别器，例如，可以将大量重影的样张输入分类器进行训练，得到分类器的参数并配置到存储设备中。存储设备对待判别图像采用这些参数进行计算后来判定是否是该类别图像。选择神经网络模型时可以采用卷积神经网络，其适合用来识别位移、缩放及其他形式扭曲不变性的二维图形，无需人工提取图像特征，能够隐式地从训练数据中学习图像特征并分类。
实际应用中，还可根据分类器的类别，来选择待判别图像的特定部分来参与判别，以提高计算速度。例如判断图像是否是眯眼照片，则只需选择人脸部分的数据计算即可，而人脸定位采用相关技术的方案即可实现，这样可以大幅减少计算量。特别是对重复照片，如果前面已有照片确定了需要判别的图像区域，则该重复照片则直接采用相同的判别图像区域进行分类判别。本发明实施例就是利用神经网络事先训练海量的各种效果不好的样张，使得智能设备能学习这些“坏”特征，获得的不同类别的分类器，从而去判别用户的图像数据库，删除重复多余的、拍摄效果差的相片，自动完成相片清理工作。特定分类器是可以将“效果不好”的相片按照不同类别进行分开，比如判断哪些是重影的，哪些是人物眯眼的等等。可选地，可以采用一个或多个差质图像分类器，即一个分类器只判断一种相片效果的好坏与否。采用多 个分类器是可以由用户选择性地对其关注的类别进行分类处理，即可单独使用也可组合使用。
在本发明的其它实施例中，一旦某张图像判定为某种效果不好的图像，则无需其他类别的判定，即可完成清理。如果是重复图像，只需其中一张图像通过所有分类器都判定为合格，则重复图像中的剩余图像可无需判定，直接进行清理。
可选地，图5为本发明实施例提供的另一种图像管理装置的结构示意图。在图3所示装置的结构基础上，本实施例提供的图像管理装置还包括：
权重设置模块60，设置为：在运算模块40确定重复图像中每张重复图像的差质权重之前，预设差质图像类别对应的差质权重；根据所述差质图像类别和对应的差质权重确定重复图像中每张重复图像的差质权重。
相应地，本实施例中运算模块40设置为当重复图像为差质图像时，确定重复图像中每张重复图像的差质权重的实现方式为：当分类模块30判断出重复图像为差质图像时，根据差质图像类别和权重设置模块60设置的与差质图像类别对应的差质权重，确定重复图像中每张重复图像的差质权重。
在本实施例中，重复图像的差质权重即本重复图像所符合的差质类型的权重之和。可以将每一差质图像类别对应一个差质权重，这个权重评分可以由用户根据个人主观感受的容忍度进行预设，例如：重影30％、曝光不足10％、人像未在预设位置20％、闭眼30％等诸如此类。当一张图像中具有重影和闭眼两个差质图像类型时，可以将重影和闭眼两个差质图像类型对应的权重进行累加，也就是说图像具有的差质特征越多，其差质权重越大，主观视觉效果越差；本发明实施例可以采用的是较为简单的评判规则，也可以采用其他公知的算法进行精确的图像差质权重计算。
可选地，图6为本发明实施例提供的图像管理装置中一种删除模块的结构示意图，如图6所示，本实施例中的删除模块50可以包括：
第一删除单元501，设置为：当运算模块40确定出重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
第二删除单元502，设置为：当运算模块40确定出重复图像中每张重复图 像的差质权重不同时，保留重复图像中差质权重最小的重复图像，删除剩余重复图像。
可选地，在本发明的一个实施例中，删除模块50还设置为：当运算模块40确定出重复图像不是差质图像时，保留其中一张重复图像，删除剩余重复图像。举例来说，当重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；当重复图像中每张重复图像的差质权重不同时，保留重复图像中差质权重最小的重复图像，删除剩余重复图像。当所有的重复图像的差质图像类型相同，例如，全部为闭眼，则直接保留一张，删除剩余的图像。当每张重复图像的差质权重不同时，例如，一张为闭眼，另外一张为闭眼和曝光不足，则第二张的质量显然比第一张更差，则保留质量最好的，删除其余照片。若重复图像不是差质图像，则仅保留其中一张重复图像，删除剩余重复图像；具体的，当所有重复图像都不具有差质图像类别时，则为合格图像，此时直接保留一张，其余删除即可。
本领域普通技术人员可以理解上述实施例的全部或部分步骤可以使用计算机程序流程来实现，所述计算机程序可以存储于一计算机可读存储介质中，所述计算机程序在相应的硬件平台上(根据系统、设备、装置、器件等)执行，在执行时，包括方法实施例的步骤之一或其组合。
可选地，上述实施例的全部或部分步骤也可以使用集成电路来实现，这些步骤可以被分别制作成一个个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。
上述实施例中的装置/功能模块/功能单元可以采用通用的计算装置来实现，它们可以集中在单个的计算装置上，也可以分布在多个计算装置所组成的网络上。
上述实施例中的装置/功能模块/功能单元以软件功能模块的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。上述提到的计算机可读取存储介质可以是只读存储器，磁盘或光盘等。
工业实用性
本发明实施例通过获取预定时间段内创建的多张图像，识别该多张图像中的重复图像，并根据预设的差质图像类别，判断上述重复图像是否为差质图像，从而在判断出复图像为差质图像时，确定每张重复图像的差质权重，并根据每张重复图像的差质权重，确定待删除的重复图像；本发明实施例解决了移动终端拍照后存储有大量重复图像的问题，实现自动删除以去除重复图像，方便用户的图像管理和移动终端内存空间的释放。

Claims (10)

  1. 一种图像管理方法，包括：
    获取预定时间段内创建的多张图像，并识别所述多张图像中的重复图像；
    根据预设的差质图像类别，判断所述重复图像是否为差质图像；
    当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重；
    根据所述重复图像中每张重复图像的差质权重，确定待删除的重复图像。
  2. 根据权利要求1所述的图像管理方法，其中，所述识别所述多张图像中的重复图像，包括：
    确定所述预定时间段内创建的所述多张图像之间的像素的均方差，将所述多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
    当所述多张图像之间的像素的均方差小于所述预设的像素的均方差阈值时，判定所述多张图像为重复图像。
  3. 根据权利要求1所述的图像管理方法，其中，所述根据所述重复图像中每张重复图像的差质权重，确定待删除的重复图像，包括：
    当所述重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
    当所述重复图像中每张重复图像的差质权重不同时，保留所述重复图像中差质权重最小的重复图像，删除剩余重复图像。
  4. 根据权利要求1所述的图像管理方法，所述方法还包括：
    当所述重复图像不是差质图像时，保留其中一张重复图像，删除剩余重复图像。
  5. 根据权利要求1所述的图像管理方法，其中，所述当所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重之前，所述方法还包括：
    预设所述差质图像类别对应的差质权重；
    所述当所述重复图像为差质图像时，确定所述重复图像中每张重复图像 的差质权重，包括：
    根据所述差质图像类别和对应的差质权重，确定所述重复图像中每张重复图像的差质权重。
  6. 一种图像管理装置，包括：
    获取模块，设置为：获取预定时间段内创建的多张图像；
    识别模块，设置为：识别所述获取模块获取的所述多张图像中的重复图像；
    分类模块，设置为：根据预设的差质图像类别，判断所述识别模块识别出的所述重复图像是否为差质图像；
    运算模块，设置为：当所述分类模块判断出所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重；
    删除模块，设置为：根据所述运算模块确定出所述重复图像中每张重复图像的差质权重，确定待删除的重复图像。
  7. 根据权利要求6所述的图像管理装置，其中，所述识别模块包括：
    确定比较单元，设置为：确定所述预定时间段内创建的所述多张图像之间像素的均方差，将所述多张图像之间的像素的均方差与预设的像素的均方差阈值进行比较；
    判定单元，设置为：当所述确定比较单元的比较结果为所述多张图像之间的像素的均方差小于所述预设的像素的均方差阈值时，判定所述多张图像为重复图像。
  8. 根据权利要求6所述的图像管理装置，其中，所述删除模块包括：
    第一删除单元，设置为：当所述运算模块确定出所述重复图像中每张重复图像的差质权重相同时，保留其中一张重复图像，删除剩余重复图像；
    第二删除单元，设置为：当所述运算模块确定出所述重复图像中每张重复图像的差质权重不同时，保留所述重复图像中差质权重最小的重复图像，删除剩余重复图像。
  9. 根据权利要求6所述的图像管理装置，其中，所述删除模块，还设置 为：当所述运算模块确定出所述重复图像不是差质图像时，保留其中一张重复图像，删除剩余重复图像。
  10. 根据权利要求6所述的图像管理装置，还包括：
    权重设置模块，设置为：在所述运算模块确定所述重复图像中每张重复图像的差质权重之前，预设所述差质图像类别对应的差质权重；
    所述运算模块设置为当所述分类模块判断出所述重复图像为差质图像时，确定所述重复图像中每张重复图像的差质权重，包括：
    当所述分类模块判断出所述重复图像为差质图像时，根据所述差质图像类别和所述权重设置模块设置的与所述差质图像类别对应的差质权重，确定所述重复图像中每张重复图像的差质权重。
PCT/CN2016/088611 2016-04-12 2016-07-05 一种图像管理方法和装置 WO2017177559A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610225927.4A CN107291781A (zh) 2016-04-12 2016-04-12 一种图像管理方法和装置
CN201610225927.4 2016-04-12

Publications (1)

Publication Number Publication Date
WO2017177559A1 true WO2017177559A1 (zh) 2017-10-19

Family

ID=60041298

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/088611 WO2017177559A1 (zh) 2016-04-12 2016-07-05 一种图像管理方法和装置

Country Status (2)

Country Link
CN (1) CN107291781A (zh)
WO (1) WO2017177559A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11527083B2 (en) * 2019-06-24 2022-12-13 Redflex Traffic Systems Pty Ltd Training a machine to recognize a motor vehicle driver using a mobile device

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110727808A (zh) * 2018-06-29 2020-01-24 华为技术有限公司 图像处理方法、装置及终端设备
WO2020073317A1 (zh) * 2018-10-12 2020-04-16 华为技术有限公司 文件管理方法及电子设备
CN111274421B (zh) * 2020-01-15 2022-03-18 平安科技（深圳）有限公司 图片数据清洗方法、装置、计算机设备和存储介质
CN113891056A (zh) * 2021-11-12 2022-01-04 盐城吉研智能科技有限公司 一种基于立体视觉技术的多摄像机图像精准分析方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1725840A (zh) * 2004-07-23 2006-01-25 三星电子株式会社 数字图像装置及其图像管理方法
CN103412753A (zh) * 2013-08-12 2013-11-27 宇龙计算机通信科技(深圳)有限公司 图片删除方法及装置
US20140028806A1 (en) * 2011-03-31 2014-01-30 Fujifilm Corporation Imaging device and imaging method
CN104954659A (zh) * 2014-03-28 2015-09-30 宇龙计算机通信科技(深圳)有限公司 一种处理图像信息的方法及装置
CN105224409A (zh) * 2015-09-30 2016-01-06 努比亚技术有限公司 一种内存的管理方法及装置
CN105472239A (zh) * 2015-11-17 2016-04-06 小米科技有限责任公司 照片处理方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1725840A (zh) * 2004-07-23 2006-01-25 三星电子株式会社 数字图像装置及其图像管理方法
US20140028806A1 (en) * 2011-03-31 2014-01-30 Fujifilm Corporation Imaging device and imaging method
CN103412753A (zh) * 2013-08-12 2013-11-27 宇龙计算机通信科技(深圳)有限公司 图片删除方法及装置
CN104954659A (zh) * 2014-03-28 2015-09-30 宇龙计算机通信科技(深圳)有限公司 一种处理图像信息的方法及装置
CN105224409A (zh) * 2015-09-30 2016-01-06 努比亚技术有限公司 一种内存的管理方法及装置
CN105472239A (zh) * 2015-11-17 2016-04-06 小米科技有限责任公司 照片处理方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11527083B2 (en) * 2019-06-24 2022-12-13 Redflex Traffic Systems Pty Ltd Training a machine to recognize a motor vehicle driver using a mobile device
US11836627B2 (en) 2019-06-24 2023-12-05 Redflex Traffic Systems Pty Ltd Training a machine to recognize a motor vehicle driver using a mobile device

Also Published As

Publication number Publication date
CN107291781A (zh) 2017-10-24

Similar Documents

Publication Publication Date Title
AU2017261537B2 (en) Automated selection of keeper images from a burst photo captured set
WO2017177559A1 (zh) 一种图像管理方法和装置
CN109284733B (zh) 一种基于yolo和多任务卷积神经网络的导购消极行为监控方法
WO2016023336A1 (zh) 照片的分类存储方法、装置、设备和计算机存储介质
Sun et al. Photo assessment based on computational visual attention model
CN108229674B (zh) 聚类用神经网络的训练方法和装置、聚类方法和装置
CN110580428A (zh) 图像处理方法、装置、计算机可读存储介质和电子设备
Villalba et al. Smartphone image clustering
CN110807759B (zh) 照片质量的评价方法及装置、电子设备、可读存储介质
JP2001325593A (ja) 自動アルバム化システムにおける重複写真検知方法
CN112597995B (zh) 车牌检测模型训练方法、装置、设备及介质
CN109325435B (zh) 基于级联神经网络的视频动作识别及定位方法
WO2022127814A1 (zh) 一种图像的显著性物体检测方法、装置、设备及存储介质
CN113010736B (zh) 一种视频分类方法、装置、电子设备及存储介质
WO2015196681A1 (zh) 一种图片处理方法及电子设备
CN108769543B (zh) 曝光时间的确定方法及装置
WO2021051382A1 (zh) 白平衡处理方法和设备、可移动平台、相机
US11816181B2 (en) Blur classification and blur map estimation
WO2021078276A1 (zh) 连拍照片获取方法、智能终端及存储介质
CN111669492A (zh) 一种终端对拍摄的数字图像进行处理的方法及终端
CN111105369B (zh) 图像处理方法、图像处理装置、电子设备和可读存储介质
CN105472405B (zh) 提醒生成方法及装置
WO2020227898A1 (zh) 图像融合方法、图像采集设备、可移动平台
CN110490027B (zh) 人脸特征提取训练方法及系统
Obrador et al. Low level features for image appeal measurement

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

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

Ref document number: 16898374

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 16898374

Country of ref document: EP

Kind code of ref document: A1