WO2017000716A2

WO2017000716A2 - Image management method and device, and terminal device

Info

Publication number: WO2017000716A2
Application number: PCT/CN2016/083634
Authority: WO
Inventors: 郑成建; 刘宇; 王勃飞; 王宁
Original assignee: 中兴通讯股份有限公司
Priority date: 2015-06-30
Filing date: 2016-05-27
Publication date: 2017-01-05
Also published as: CN106326908A; WO2017000716A3

Abstract

An image management method and device, and a terminal device. The method comprises: classifying images in a terminal device, and obtaining classified images, the classified images including type labels of image categories; obtaining degrees of similarity of images having a same type label, grouping images having a same type label according to the degrees of similarity, and obtaining at least one group of images, the degree of similarity of any two images within each group of images being greater than a preset threshold; within each group of images, selecting a local image to be saved in the terminal device, obtaining a difference image of each remaining image of the group and the local image, storing the difference images in a cloud, and simultaneously deleting the remaining images of the group.

Description

图片的管理方法、装置及终端设备Picture management method, device and terminal device

技术领域Technical field

本申请涉及图像处理技术领域，例如涉及一种图片的管理方法、装置及终端设备。The present application relates to the field of image processing technologies, and relates to a method, an apparatus, and a terminal device for managing a picture, for example.

背景技术Background technique

随着终端设备(如手机、平板电脑等)性能越来越强大，拍照的功能越来越出色，终端设备中的照片越来越多。大部分用户终端设备中都存在大量相似的照片，这些相似照片有的是由于用户在拍照时为取得一张最好的效果而故意多拍的，有的是使用连拍功能所产生的，有的是使用美化软件等保留的副本。而随着硬件配置的升级，像素分辨率越来越高，单张照片的存储量越来越大，久而久之这些照片就会占用大量的手机内存。As the performance of terminal devices (such as mobile phones, tablets, etc.) becomes more and more powerful, the functions of photographing are getting better and better, and the photos in the terminal devices are more and more. A large number of similar photos exist in most user terminal devices. Some of these similar photos are deliberately taken by the user to obtain a best effect when taking a picture. Some are generated by using the continuous shooting function, and some are using landscaping software. A copy of the reservation. With the upgrade of the hardware configuration, the pixel resolution is getting higher and higher, and the storage capacity of a single photo is getting larger and larger. Over time, these photos will occupy a large amount of mobile phone memory.

发明内容Summary of the invention

本发明实施例提供一种图片的管理方法、装置及终端设备，解决终端设备中相似图片占用大量宝贵内存的问题。The embodiments of the present invention provide a method, a device, and a terminal device for managing a picture, which solve the problem that a similar picture in the terminal device occupies a large amount of valuable memory.

本发明实施例提供了一种图片的管理方法，包括：An embodiment of the present invention provides a method for managing a picture, including:

对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签；Performing classification processing on the picture in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category;

获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的任意两张图片之间的相似度大于预设阈值；Obtaining similarity between pictures having the same class label, and grouping the pictures with the same class label according to the similarity to obtain at least one set of pictures, wherein any two pictures in each set of pictures The similarity between the two is greater than a preset threshold;

在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像存储到云端，同时删除该组剩余的图片。Selecting a local image to be saved in the terminal device in each group of pictures, and obtaining a difference image between each of the remaining pictures of the group and the local picture, storing the difference image in the cloud, and deleting the group The remaining pictures.

其中，上述的图片的管理方法，还包括：The above method for managing pictures further includes:

从云端获取所述差值图像，并将所述差值图像与对应的本地图片相加，恢复被删除的图片。The difference image is acquired from the cloud, and the difference image is added to the corresponding local picture to restore the deleted picture.

其中，所述对终端设备中的图片进行分类处理，得到分类后的图片，包括： The classifying the pictures in the terminal device to obtain the classified pictures, including:

检测所述终端设备的图片中是否包含人脸信息；Detecting whether the face information is included in the picture of the terminal device;

若包含人脸信息，则将所述图片划分为人物类图片，若不包含人脸信息，将所述图片划分为风景类图片。If the face information is included, the picture is divided into a character type picture, and if the face information is not included, the picture is divided into a landscape picture.

其中，所述获取具有相同类标签的图片之间的相似度，包括：The obtaining the similarity between the pictures with the same class label includes:

根据图片的类标签，提取终端设备中每张图片的特征信息；Extracting feature information of each picture in the terminal device according to the class label of the picture;

根据所述特征信息，获取具有相同类标签的图片之间的相似度。Obtaining similarities between pictures having the same class label according to the feature information.

其中，所述根据图片的类标签，提取每张图片的特征信息，包括：The extracting the feature information of each image according to the class label of the image, including:

若所述图片的类标签表明所述图片为人物类图片，则提取图片的特征信息包括图片的颜色直方图及人脸区域的轮廓特征；If the class label of the picture indicates that the picture is a character class picture, the feature information of the extracted picture includes a color histogram of the picture and a contour feature of the face area;

若所述图片的类标签表明所述图片为风景类图片，则提取图片的特征信息包括图片的颜色直方图。If the class label of the picture indicates that the picture is a landscape picture, the feature information of the extracted picture includes a color histogram of the picture.

其中，所述根据所述特征信息，获取具有相同类标签的图片之间的相似度，包括：The obtaining the similarity between the pictures having the same class label according to the feature information, including:

在所述终端设备的图片中选取一张未进行分组的第一图片，并获取所述第一图片的类标签；Selecting a first picture that is not grouped in the picture of the terminal device, and acquiring a class label of the first picture;

若所述类标签表明所述第一图片为人物类图片，则在所述终端设备未进行分组的图片中选取类标签为人物类图片的第二图片，并根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度；If the class label indicates that the first picture is a character class picture, selecting a second picture of the character class picture in the picture that is not grouped by the terminal device, and according to the color histogram and the face of the picture Calculating a similarity between the first picture and the second picture by using a contour feature of the area;

若所述类标签表明所述第一图片为风景类图片，则在所述终端设备未进行分组的图片中选取类标签为风景类图片的第三图片，并根据图片的颜色直方图，计算所述第三图片与所述第一图片的相似度。If the class label indicates that the first picture is a landscape picture, select a third picture whose class label is a landscape picture in a picture that is not grouped by the terminal device, and calculate a location according to a color histogram of the picture. The similarity between the third picture and the first picture is described.

其中，所述根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度，包括：The calculating the similarity between the first picture and the second picture according to the color histogram of the picture and the contour feature of the face area, including:

判断所述第一图片与所述第二图片中的人脸个数是否相同；Determining whether the number of faces in the first picture and the second picture are the same;

若相同，则通过公式S＝r1*P+r2*H计算所述第一图片和所述第二图片的相似度；If the same, the similarity between the first picture and the second picture is calculated by the formula S=r1*P+r2*H;

其中，S表示相似度，r1和r2为预设系数，且r1+r2＝1，P表示人脸区域的轮廓特征的匹配度，H表示颜色直方图相似度。Where S represents the similarity, r1 and r2 are preset coefficients, and r1+r2=1, P represents the matching degree of the contour feature of the face region, and H represents the color histogram similarity.

其中，所述在每组图片中选取出需要保存在所述终端设备中的本地图片的步骤包括：The step of selecting a local picture that needs to be saved in the terminal device in each group of pictures includes:

获取每组图片的图片质量； Get the image quality of each set of images;

在每组图片中，选取图片质量最高的一张图片作为所述本地图片。In each set of pictures, one picture with the highest picture quality is selected as the local picture.

本发明实施例还提供了一种图片的管理装置，包括：The embodiment of the invention further provides a picture management device, including:

分类模块，设置为对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签；a classification module, configured to classify a picture in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category;

分组模块，设置为获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的两张图片之间的相似度大于预设阈值；a grouping module, configured to acquire a similarity between pictures having the same class label, and perform grouping processing on the pictures with the same class label according to the similarity to obtain at least one group of pictures, where each group of pictures The similarity between the two pictures is greater than a preset threshold;

处理模块，设置为在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像存储到云端，同时删除该组剩余的图片。a processing module, configured to select a local picture that needs to be saved in the terminal device in each group of pictures, and obtain a difference image between each picture of the group and the local picture, and store the difference image in the cloud , while deleting the remaining pictures of the group.

其中，上述图片的管理装置，还包括：The management device of the above picture further includes:

恢复模块，设置为从云端获取所述差值图像，并将所述差值图像与对应的本地图片相加，恢复被删除的图片。And a recovery module, configured to acquire the difference image from the cloud, and add the difference image to the corresponding local image to recover the deleted picture.

其中，所述分类模块包括：The classification module includes:

检测子模块，设置为检测所述终端设备的图片中是否包含人脸信息；a detecting submodule, configured to detect whether the face information is included in the picture of the terminal device;

划分子模块，设置为若所述检测模块检测到所述终端设备的图片中包含人脸信息，则将所述图片划分为人物类图片，否则，将所述图片划分为风景类图片。And dividing the sub-module, the method is configured to divide the picture into a character-like picture if the detection module detects that the picture of the terminal device includes face information, and otherwise divide the picture into a landscape-like picture.

其中，所述分组模块包括：The grouping module includes:

提取子模块，设置为根据图片的类标签，提取终端设备中每张图片的特征信息；Extracting a sub-module, configured to extract feature information of each picture in the terminal device according to the class label of the picture;

获取子模块，设置为根据所述特征信息，获取具有相同类标签的图片之间的相似度。The obtaining submodule is configured to acquire the similarity between the pictures having the same class label according to the feature information.

其中，所述提取子模块包括：The extracting submodule includes:

第一获取单元，设置为若所述图片的类标签表明所述图片为人物类图片，则获取所述图片的颜色直方图及人脸区域的轮廓特征；a first acquiring unit, configured to acquire a color histogram of the picture and a contour feature of the face area if the class label of the picture indicates that the picture is a character class picture;

第二获取单元，设置为若所述图片的类标签表明所述图片为风景类图片，则获取所述图片的颜色直方图。The second obtaining unit is configured to obtain a color histogram of the picture if the class label of the picture indicates that the picture is a landscape picture.

其中，所述获取子模块包括：The obtaining submodule includes:

第三获取单元，设置为在所述终端设备的图片中选取一张未进行分组的第一图片，并获取所述第一图片的类标签； a third acquiring unit, configured to: select a first picture that is not grouped in the picture of the terminal device, and acquire a class label of the first picture;

第一计算单元，设置为若所述类标签表明所述第一图片为人物类图片，则在所述终端设备未进行分组的图片中选取类标签为人物类图片的第二图片，并根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度；a first calculating unit, configured to: if the class label indicates that the first picture is a character class picture, select a second picture of the character class picture in the picture that is not grouped by the terminal device, and according to the picture a color histogram and a contour feature of the face region, and calculating a similarity between the first picture and the second picture;

第二计算单元，设置为若所述类标签表明所述第一图片为风景类图片，则在所述终端设备未进行分组的图片中选取类标签为风景类图片的第三图片，并根据图片的颜色直方图，计算所述第三图片与所述第一图片的相似度。a second calculating unit, configured to: if the class label indicates that the first picture is a landscape picture, select a third picture whose class label is a landscape picture in a picture that is not grouped by the terminal device, and according to the picture a color histogram, calculating a similarity between the third picture and the first picture.

其中，所述第一计算单元包括：The first computing unit includes:

判断子单元，设置为判断所述第一图片与所述第二图片中的人脸个数是否相同；a determining subunit, configured to determine whether the number of faces in the first picture and the second picture are the same;

计算子单元，设置为若所述第一图片与所述第二图片中的人脸个数相同，则通过公式S＝r1*P+r2*H计算所述第一图片和所述第二图片的相似度；Calculating a subunit, configured to calculate the first picture and the second picture by using a formula S=r1*P+r2*H if the number of faces in the first picture and the second picture are the same Similarity

其中，所述处理模块包括：The processing module includes:

第四获取子模块，设置为获取每组图片的图片质量；The fourth obtaining submodule is configured to obtain a picture quality of each group of pictures;

选取子模块，设置为在每组图片中，选取图片质量最高的一张图片作为所述本地图片。The sub-module is selected, and in each set of pictures, one picture with the highest picture quality is selected as the local picture.

本发明实施例还提供了一种终端设备，包括如上所述的图片的管理装置。The embodiment of the invention further provides a terminal device, which comprises the management device of the picture as described above.

本发明实施例还提供一种非易失性计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行上述方法。Embodiments of the present invention also provide a non-transitory computer readable storage medium storing computer executable instructions for performing the above method.

本发明实施例的图片的管理方法，对终端设备中的图片进行分类处理，并将归属于同一类别、且相似度大于预设阈值的图片划分为一组，在每组图片中选取出本地图片保存在所述终端设备中，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像信息存储到云端，同时删除该组剩余的图片，极大的节约了终端设备的存储空间。另外，本发明实施例中，将图片存储到云端时，只对差值图像进行编码传输，而由于每组图片相似性很大，差值图像编码后码流相对较小，比传输原图更节约带宽，且由于本发明实施例在云端仅存储差值图像，因此当存储在云端的差值图像被解码后也无法看出原图片的内容，有效防止因***漏洞或其他原因造成的图片泄露。 The method for managing a picture in the embodiment of the present invention classifies a picture in a terminal device, and divides the pictures belonging to the same category and whose similarity is greater than a preset threshold into a group, and selects a local picture in each group of pictures. Saving in the terminal device, and acquiring a difference image between each picture of the group and the local picture, storing the difference image information in the cloud, and deleting the remaining pictures of the group, thereby greatly saving the terminal The storage space of the device. In addition, in the embodiment of the present invention, when the picture is stored in the cloud, only the difference image is encoded and transmitted, and since the similarity of each group of pictures is large, the code stream after the difference image encoding is relatively small, which is more than the original picture. The bandwidth is saved, and since the difference image is stored in the cloud in the embodiment of the present invention, the content of the original image cannot be seen after the difference image stored in the cloud is decoded, thereby effectively preventing image leakage caused by system loopholes or other reasons. .

附图说明DRAWINGS

图1表示本发明实施例的图片的管理方法的第一工作流程图；1 is a first working flowchart of a method for managing a picture according to an embodiment of the present invention;

图2表示本发明实施例的图片的管理方法的第二工作流程图；2 is a second working flowchart of a method for managing a picture according to an embodiment of the present invention;

图3表示本发明实施例的图片的管理装置的结构框图；3 is a block diagram showing the structure of a picture management apparatus according to an embodiment of the present invention;

图4表示本发明实施例的终端设备的硬件结构示意图。FIG. 4 is a schematic diagram showing the hardware structure of a terminal device according to an embodiment of the present invention.

实施方式Implementation

为使本申请要解决的技术问题、技术方案清楚，下面将结合可选实施例及附图进行描述。In order to make the technical problems and technical solutions to be solved by the present application clear, the following description will be made in conjunction with the optional embodiments and the accompanying drawings.

相关的图片管理方法可以包括三个方面，一是研究连拍照片的压缩存储，利用连拍照片重复像素多的特点进行编码来减少照片的存储空间，但压缩后还是要占用终端设备的存储空间，对终端设备存储空间的节省有限。二是应用图像处理技术进行择优选取，在一组相似照片中只保留一张，删除其余的相似照片，它的缺点是无法保留所有照片，删除后无法恢复，后续想再回顾以往拍摄的照片，只能看到保留的那一张。三是采用云存储，直接在云端备份所有的照片，当照片的容量超过云端的预定容量时，用户需有偿使用云端，且这中存储照片的方式存在照片泄露的风险。The related picture management method can include three aspects. One is to study the compressed storage of continuous shooting photos, and to use the continuous shooting photos to repeat the characteristics of multiple pixels to reduce the storage space of the photos, but still need to occupy the storage space of the terminal device after compression. There is limited savings on the storage space of the terminal device. The second is to use image processing technology to select and choose. Only one of a group of similar photos is kept, and the rest of similar photos are deleted. The disadvantage is that all photos cannot be retained, and cannot be restored after deletion. Later, I want to review the photos taken in the past. Only see the one that was retained. The third is to use cloud storage to back up all photos directly in the cloud. When the capacity of the photo exceeds the predetermined capacity of the cloud, the user needs to use the cloud for a fee, and there is a risk of photo leakage in the way the photo is stored.

因此，本发明实施例提供了一种图片的管理方法、装置及终端设备，解决了终端设备中相似图片占用大量宝贵内存的问题。Therefore, the embodiment of the present invention provides a method, a device, and a terminal device for managing a picture, which solves the problem that a similar picture in the terminal device occupies a large amount of valuable memory.

本发明实施例的图片的管理方法，如图1所示，包括：A method for managing a picture in the embodiment of the present invention, as shown in FIG. 1, includes:

步骤S11：对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签。Step S11: Perform classification processing on the pictures in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category.

其中，在对终端设备中的图片进行分类时，可将图片分为人物类、风景类等，不同类别的图片具有不同的类标签。When the pictures in the terminal device are classified, the pictures may be divided into a character class, a landscape class, and the like, and different types of pictures have different class tags.

步骤S12：获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的两张图片之间的相似度大于预设阈值。Step S12: Acquire similarity between pictures having the same class label, and perform grouping processing on the pictures with the same class label according to the similarity to obtain at least one set of pictures, wherein two of each set of pictures The similarity between pictures is greater than a preset threshold.

步骤S13：在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像存储到云端，同时删除该组剩余的图片。Step S13: selecting a local picture that needs to be saved in the terminal device in each group of pictures, and acquiring a difference image between each picture of the group and the local picture, and storing the difference image in the cloud, Delete the remaining pictures of the group.

本发明实施例的图片的管理方法，对终端设备中的图片进行分类处理，并将归属于同一类别、且相似度大于预设阈值的图片划分为一组，在每组图片中选取出本地图片保存在所述终端设备中，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像信息存储到云端，同时删除该组剩余的图片，极大的节约了终端设备的存储空间。另外，本发明实施例中，将图片存储到云端时，只对差值图像进行编码传输，而由于每组图片相似性很大，差值图像编码后码流相对较小，比传输原图更节约带宽，且由于本发明实施例在云端仅存储差值图像，因此当存储在云端的差值图像被解码后也无法看出原图像内容，有效防止因***漏洞或其他原因造成的图片泄露。The method for managing pictures in the embodiment of the present invention performs classification processing on pictures in the terminal device, and Dividing a picture belonging to the same category and having a similarity greater than a preset threshold into a group, selecting a local picture in each group of pictures to be saved in the terminal device, and acquiring each remaining picture of the group and the local The difference image of the picture stores the difference image information in the cloud, and deletes the remaining pictures of the group, which greatly saves the storage space of the terminal device. In addition, in the embodiment of the present invention, when the picture is stored in the cloud, only the difference image is encoded and transmitted, and since the similarity of each group of pictures is large, the code stream after the difference image encoding is relatively small, which is more than the original picture. The bandwidth is saved, and since the difference image is stored in the cloud in the embodiment of the present invention, the original image content cannot be seen after the difference image stored in the cloud is decoded, thereby effectively preventing image leakage caused by system loopholes or other reasons.

可选地，所述步骤S11包括：Optionally, the step S11 includes:

步骤S111：检测所述终端设备的图片中是否包含人脸信息；Step S111: detecting whether the face information is included in the picture of the terminal device;

步骤S112：若包含人脸信息，则将所述图片划分为人物类图片，若不包含人脸信息，将所述图片划分为风景类图片。Step S112: If the face information is included, the picture is divided into a character type picture, and if the face information is not included, the picture is divided into a landscape picture.

在本发明实施例中，可通过对每张图片进行人脸检测，来将图片划分为人物类图片和风景类图片，可选地，人脸检测的方法可以采用基于肤色的检测方法，将图片转化到颜色空间YCbCr，其中，Y表示图片的亮度分量，Cb表示图片的蓝色色度分量，Cr表示图片的红色色度分量。利用肤色投影到CbCr平面的聚类性来检测图片中的皮肤区域，其中Cb取值区间为[90，128]，Cr取值区间为[110，190]，该肤色区间的阈值可经过多次测试进行调整。In the embodiment of the present invention, the picture may be divided into a character class picture and a landscape class picture by performing face detection on each picture. Optionally, the face detection method may adopt a skin color based detection method to display the picture. Converted to the color space YCbCr, where Y represents the luminance component of the picture, Cb represents the blue chrominance component of the picture, and Cr represents the red chrominance component of the picture. The skin area in the picture is detected by the clustering of the skin color projection to the CbCr plane, wherein the Cb value interval is [90, 128], and the Cr value interval is [110, 190], and the skin color interval threshold can be repeated multiple times. The test is adjusted.

对上述检测结果得到的二值化图像进行形态学运算，先腐蚀排除干扰点，再膨胀填充空洞点。The morphological operation is performed on the binarized image obtained by the above detection result, and the interference point is first etched and then the void point is expanded and expanded.

可选地，通过区域标记技术提取形态学运算后的二值化图像上的人脸区域，获取每个区域的外接矩形的长宽信息。根据生物学方面的知识，人脸长宽比相对固定，一般在0.8-1.25之间，通过这个特征，排除误检区域，统计当前图片中的人脸个数。Optionally, the face region on the binarized image after the morphological operation is extracted by the region labeling technique, and the length and width information of the circumscribed rectangle of each region is obtained. According to the biological knowledge, the aspect ratio of the face is relatively fixed, generally between 0.8 and 1.25. Through this feature, the false detection area is eliminated, and the number of faces in the current picture is counted.

可选地，所述步骤S12中获取具有相同类标签的图片之间的相似度，包括：Optionally, obtaining the similarity between the pictures having the same class label in the step S12, including:

步骤S121：根据图片的类标签，提取终端设备中每张图片的特征信息；Step S121: Extract feature information of each picture in the terminal device according to the class label of the picture;

步骤S122：根据所述特征信息，获取具有相同类标签的图片之间的相似度。Step S122: Acquire similarities between pictures having the same class label according to the feature information.

其中，所述步骤S121可包括：The step S121 may include:

若所述图片的类标签表明所述图片为人物类图片，则提取所述图片特征信息包括图片的颜色直方图及人脸区域的轮廓特征；If the class label of the picture indicates that the picture is a character class picture, extracting the picture feature information includes a color histogram of the picture and a contour feature of the face area;

若所述图片的类标签表明所述图片为风景类图片，则提取所述图片的特征信息包括图片的颜色直方图。If the class label of the picture indicates that the picture is a landscape picture, extracting features of the picture The information includes the color histogram of the picture.

在本发明实施例中，提取图片中人脸区域的轮廓特征可以包括：将人脸区域统一缩放为L*L像素，L为预设值，其中，L表示人脸区域的大小，在不同图片中人脸区域的大小可能不一致，在本实施例中将不同图片中的人脸区域统一缩放为像素大小为L*L。采用sobel算子对缩放后的人脸区域图像进行边缘检测，应用“大律法”求出检测结果图像的二值化阈值，对检测结果图像进行二值化，二值化后的结果作为该人脸区域的轮廓特征矩阵保存。可选地，预设值L为80，L的取值的确定可经过多次测试得到。In the embodiment of the present invention, extracting the contour feature of the face region in the image may include: uniformly scaling the face region to L*L pixels, where L is a preset value, where L represents the size of the face region, in different images. The size of the face area may be inconsistent. In this embodiment, the face area in different pictures is uniformly scaled to a pixel size of L*L. Edge detection is performed on the scaled face region image by using the sobel operator, and the binarization threshold of the detection result image is obtained by using the "big law method", and the detection result image is binarized, and the result of the binarization is used as the result. The contour feature matrix of the face area is saved. Optionally, the preset value L is 80, and the determination of the value of L can be obtained through multiple tests.

在本发明实施例中，两个图片之间的颜色直方图的相似性可以采用巴氏距离来确定。In an embodiment of the invention, the similarity of the color histograms between the two pictures may be determined using a Pap singer distance.

其中，所述步骤S122可包括：The step S122 may include:

可选地，所述根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度，包括：Optionally, calculating the similarity between the first picture and the second picture according to the color histogram of the picture and the contour feature of the face area, including:

其中，S表示相似度，r1和r2为预设系数，且r1+r2＝1，P表示人脸区域的轮廓特征的匹配度，H表示颜色直方图相似度。可选地，预设系数r1、r2为0.4、0.6，预设系数r1、r2取值的确定可经过多次测试得到。Where S represents the similarity, r1 and r2 are preset coefficients, and r1+r2=1, P represents the matching degree of the contour feature of the face region, and H represents the color histogram similarity. Optionally, the preset coefficients r1 and r2 are 0.4 and 0.6, and the determination of the values of the preset coefficients r1 and r2 can be obtained through multiple tests.

在本发明实施例中，计算人脸区域的轮廓特征的匹配度P包括：将图片A和图片B的人脸区域轮廓特征矩阵进行异或运算，统计运算结果中1的个数M，M的值越小则表示图片A和图片B越相似；将M的值归一化后便可得到人脸区域的轮廓特征的匹配度P，P＝M/(L*L)。 In the embodiment of the present invention, calculating the matching degree P of the contour feature of the face region includes: performing an exclusive OR operation on the contour feature matrix of the face region of the picture A and the picture B, and counting the number of the M in the statistical operation result M, M The smaller the value, the more similar the picture A and the picture B are; the normalized value of M can be used to obtain the matching degree P of the contour feature of the face region, P=M/(L*L).

可选地，根据图片的颜色直方图，计算所述第三图片与所述第一图片的相似度可包括：Optionally, calculating the similarity between the third picture and the first picture according to the color histogram of the picture may include:

将所述第三图片和所述第一图片的颜色直方图的相似度作为所述第三图片和所述第一图片的相似度。Using the similarity of the color histogram of the third picture and the first picture as the similarity of the third picture and the first picture.

根据上述方法得到具有相同类标签的图片之间的相似度S后，判断具有相同类标签的两张图片之间的相似度是否大于预设阈值T1，并在大于预设阈值时，划分为同一组图片，其中，预设阈值T1可选为0.8，该预设阈值T1取值的确定可经过多次测试得到。同时，在本发明实施例中，赋予每组图片一个组标签，组标签的设置可以包含当前组的图片类别、该组中相似图片的个数以及唯一的序号，例如用“人物_6_20150326H000123”来表示该组为人物类图片，且该组中共有6张相似的图片，最后的一串即为该组图片的唯一序号，可选地，该序号中可涵盖图片拍摄时间、拍摄位置等信息。After obtaining the similarity S between the pictures having the same class label according to the foregoing method, it is determined whether the similarity between the two pictures having the same class label is greater than a preset threshold T1, and is greater than the preset threshold, and is divided into the same The group picture, wherein the preset threshold T1 can be selected as 0.8, and the determination of the value of the preset threshold T1 can be obtained through multiple tests. Meanwhile, in the embodiment of the present invention, each group of pictures is given a group label, and the setting of the group label may include the picture category of the current group, the number of similar pictures in the group, and a unique serial number, for example, “person_6_20150326H000123” The group is a character-like picture, and there are 6 similar pictures in the group, and the last string is the unique serial number of the group of pictures. Optionally, the serial number can cover information such as picture capturing time and shooting position.

可选地，所述步骤S13中在每组图片中选取出需要保存在所述终端设备中的本地图片包括：Optionally, the selecting, in the step S13, the local pictures that need to be saved in the terminal device in each group of pictures includes:

获取每组图片的图片质量；Get the image quality of each set of images;

当然，在本发明实施例中，也可以根据用户需求在每组中选取出任意一张或者多张图片作为本地图片，保存在终端设备中。Of course, in the embodiment of the present invention, any one or more pictures may be selected as a local picture in each group according to user requirements, and stored in the terminal device.

其中，获取每组图片中每张图片的图片质量的步骤可以具体包括：The step of obtaining the picture quality of each picture in each group of pictures may specifically include:

步骤S131：计算当前图片的灰度图像；Step S131: calculating a grayscale image of the current picture;

步骤S132：计算步骤S131得到的灰度图像的信息熵F；Step S132: calculating the information entropy F of the grayscale image obtained in step S131;

步骤S133：计算当前图片的色彩系数Color；Step S133: calculating a color coefficient of the current picture Color;

步骤S134：若当前图片为人物类图片，计算人脸区域对应的灰度图像的对比度Contrast；Step S134: If the current picture is a character class picture, calculate a contrast Contrast of the gray image corresponding to the face area;

步骤S135：根据当前图片的类标签以及信息熵F、色彩系数Color和/或对比度Contrast计算图片质量。Step S135: Calculate the picture quality according to the class label of the current picture and the information entropy F, the color coefficient Color, and/or the contrast Contrast.

在步骤S132中，图像的信息熵

式中：K为灰度值，p_K为灰度值K在该灰度图像中出现的概率，N为图像的灰度等级。In step S132, the information entropy of the image

Where: K is the gray value, p _K is the probability that the gray value K appears in the gray image, and N is the gray level of the image.

在步骤S133中，图像的色彩系数

式中：

其中，σ表示图像色彩系数的标准差，μ表示图像色彩系数的均值，rg、yb是图像RGB色彩空间的一个简单对应色彩空间： rg＝R-G，yb＝0.5*(R+G)-B，其中，R表示RGB色彩空间中的红色分量值；G表示RGB色彩空间中的绿色分量值；B表示RGB色彩空间中的蓝色分量值。In step S133, the color coefficient of the image

In the formula:

Where σ represents the standard deviation of the image color coefficient, μ represents the mean value of the image color coefficient, and rg and yb are a simple corresponding color space of the image RGB color space: rg=RG, yb=0.5*(R+G)-B, Where R represents the red component value in the RGB color space; G represents the green component value in the RGB color space; B represents the blue component value in the RGB color space.

在步骤S134中，人脸区域的对比度contrast可采用方差对比度

式中：L表示人脸区域的大小，I表示当前位置人脸的亮度，

是整个人脸区域的平均亮度。In step S134, the contrast contrast of the face region can adopt the variance contrast

Where: L represents the size of the face area, and I represents the brightness of the face at the current position.

It is the average brightness of the entire face area.

在获取到每组图片的图片质量后，对每组中的图片按质量从高到低进行排序，标记每组中质量最高的图片，赋予一个ID。After obtaining the picture quality of each group of pictures, the pictures in each group are sorted by quality from high to low, and the highest quality picture in each group is marked, and an ID is assigned.

可选地，对每组中的图片按质量从高到低进行排序的步骤可包括：首先对图像的信息熵F按照从大到小的顺序进行排序；然后对排序后的图像赋予不同的分值，例如信息熵F最大的图像分值为100，其余图像按比例取值[0，100]。对色彩系数Color进行相同的操作，先排序，后赋予分值。若当前图片为风景类图片，则风景类图片的图片质量的系数Q＝F_d+Color_d，式中F_d、Color_d分别表示信息熵F和色彩系数Color对应的分值；若当前图片为人物类图片，则继续对人脸区域的对比度Contrast进行排序，同样映射取值[0，100]，此时人物类图片的图片质量的系数Q＝F_d+Color_d+Contrast_d，Contrast_d表示人脸区域的对比度Contrast对应的分数值。根据计算得到的当前图片的图片质量系数，可获取该当前图片的图片质量，其中，图片质量的系数Q越大，表明该当前图片的质量越好。Optionally, the step of sorting the pictures in each group according to the quality from high to low may include: first ordering the information entropy F of the images in descending order; and then assigning different points to the sorted images. The value, for example, the maximum image entropy F is 100, and the remaining images are proportional [0, 100]. Perform the same operation on the color coefficient Color, sort first, and then assign a score. If the current picture is a landscape picture, the picture quality coefficient of the landscape picture is Q=F _d +Color _d , where F _d and Color _d respectively represent the score corresponding to the information entropy F and the color coefficient Color; if the current picture is The character class picture continues to sort the contrast Contrast of the face region, and the mapping takes the value [0, 100]. At this time, the coefficient of the picture quality of the character class picture is Q=F _d +Color _d +Contrast _d , Contrast _d indicates The score of the Contrast corresponding to the contrast of the face area. According to the calculated picture quality coefficient of the current picture, the picture quality of the current picture may be obtained, where the coefficient Q of the picture quality is larger, indicating that the quality of the current picture is better.

可选地，本发明实施例的图片的管理方法，还包括：Optionally, the method for managing a picture in the embodiment of the present invention further includes:

本发明实施例的图片的管理方法，将图片的差值图像存储到云端，用户无需为节省终端设备的存储空间而删除图片，所有图片可以随时回溯、浏览；且将图片的差值图像存储到云端时，只对差值图像进行编码传输，而由于每组图片相似性很大，差值图像编码后码流相对较小，比传输原图更节约带宽。另外，将原始图像(存储在终端设备中)和差值图像(存储在云端)分开存储，只需对原始图像采取较高的保护措施，存储在云端的差值图像解码后无法看出图片内容，有效防止因***漏洞或其他原因造成的图片泄露。The image management method of the embodiment of the present invention stores the difference image of the image in the cloud, and the user does not need to delete the image to save the storage space of the terminal device, and all the images can be backtracked and browsed at any time; and the difference image of the image is stored to In the cloud, only the difference image is encoded and transmitted, and since the similarity of each group of pictures is large, the code stream after the difference image encoding is relatively small, which saves bandwidth more than the original picture. In addition, the original image (stored in the terminal device) and the difference image (stored in the cloud) are stored separately, and only a high protection measure is taken on the original image, and the image content cannot be seen after decoding the difference image stored in the cloud. To effectively prevent image leaks caused by system vulnerabilities or other reasons.

以上分别就图片的管理方法的步骤做出了解释说明，下面对图片的管理方法的整体流程进行说明。The above explains the steps of the picture management method separately, and the overall flow of the picture management method is explained below.

如图2所示，包括： As shown in Figure 2, it includes:

步骤S21：对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签。Step S21: Perform classification processing on the pictures in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category.

可选地，可将图片分为人物类图片和风景类图片。Optionally, the picture can be divided into a character class picture and a landscape class picture.

步骤S22：根据图片的类别，提取图片的特征信息。Step S22: Extract feature information of the picture according to the category of the picture.

步骤S23：根据图片的特征信息，获取具有相同标签的图片之间的相似度。Step S23: Acquire similarities between pictures having the same label according to the feature information of the picture.

步骤S24：根据所述相似度对所述具有相同类标签的图片进行分组处理，每组图片中的任意两张图片之间的相似度大于预设阈值。Step S24: Perform grouping processing on the pictures with the same class label according to the similarity, and the similarity between any two pictures in each group of pictures is greater than a preset threshold.

步骤S25：计算每组图片的质量，并标记质量最高的图片，同时获取该图片与该组剩余图片的差值图像。Step S25: Calculate the quality of each group of pictures, and mark the picture with the highest quality, and obtain the difference image of the picture and the remaining pictures of the group.

步骤S26：对差值图像进行编码，并将差值图像的编码传输到云端。Step S26: Encoding the difference image and transmitting the code of the difference image to the cloud.

步骤S27：在终端设备中只保存质量最高的图片，删除该组剩余的图片。Step S27: Only the highest quality picture is saved in the terminal device, and the remaining pictures of the group are deleted.

步骤S28：当需要查看保留的质量最高图片以外的图片时，从云端获取相应的差值图像码流，解码并与本组质量最高的图片相加还原被删除的图片。Step S28: When it is necessary to view the picture other than the highest quality picture, the corresponding difference image code stream is obtained from the cloud, decoded and added to the picture with the highest quality of the group to restore the deleted picture.

可选地，首先获取终端设备中当前标记的本地图片所在组的组标签，该组标签中包括图片类别、该组中相似图片的个数以及唯一的序号，解析组标签获得当前组相似图片的个数；然后根据组标签的唯一序号从云端获取该组图片对应的全部的差值图像码流；最后，在终端设备解码差值图像的码流，并将解码结果与当前本地图片相加获得对应的被删除的图片原图，若终端设备中无法找到当前本地图片，则只返回差值图像，并提示错误。Optionally, the group label of the group in which the local picture currently marked in the terminal device is located is first obtained, where the group label includes a picture category, a number of similar pictures in the group, and a unique serial number, and the parsing group label obtains a similar picture of the current group. And then obtaining all the difference image code streams corresponding to the group of pictures from the cloud according to the unique serial number of the group label; finally, decoding the code stream of the difference image at the terminal device, and adding the decoding result to the current local picture Corresponding deleted picture original picture, if the current local picture cannot be found in the terminal device, only the difference image is returned and an error is indicated.

本发明实施例提供的的图片管理方法，应用图像分析技术，首先对终端设备中的图片进行分类，然后提取分类后的图片的特征，并利用基于这些特征定义的相似度量函数计算特征之间的相似性，将相似的图片标记为一组；再通过设计数码图片质量评价方案，选取每组中的最优图片；最后通过对差值图像编码，保存到云存储的方式，删除终端设备端的相似图片。而当用户需要浏览图片时，只需要通过相应的逆过程即可实现浏览全部图片。本发明实施例采用图像分析技术，并将它与云存储相结合，解决了终端设备相似图片多，整理麻烦，占用终端设备的存储空间大的问题，且所有图片可以随时回溯、浏览，无需在删除图片和节省内存间进行选择，同时采用存储差值图像的方式，提高图片的保密度，保护用户隐私。The image management method provided by the embodiment of the present invention applies image analysis technology to first classify pictures in the terminal device, and then extracts features of the classified pictures, and calculates similar features between the features based on the similarity measurement functions defined by the features. Similarity, mark similar pictures as a group; then select the best picture in each group by designing digital picture quality evaluation scheme; finally, by encoding the difference image, save to cloud storage, delete the similarity of terminal equipment image. When the user needs to browse the picture, only the corresponding reverse process can be used to browse all the pictures. The embodiment of the invention adopts the image analysis technology and combines it with the cloud storage to solve the problem that the terminal device has many similar pictures, the arrangement is troublesome, and the storage space of the terminal device is occupied, and all the pictures can be backtracked and browsed at any time without Select pictures and save memory to select between them, and use the method of storing difference images to improve the density of images and protect user privacy.

本发明实施例还提供了一种图片的管理装置，如图3所示，包括：The embodiment of the invention further provides a picture management device, as shown in FIG. 3, comprising:

分类模块31，设置为对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签；The classification module 31 is configured to classify the pictures in the terminal device to obtain a classified picture. a slice, the classified picture includes a class label indicating the picture category;

分组模块32，设置为获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的任意两张图片之间的相似度大于预设阈值；The grouping module 32 is configured to acquire the similarity between the pictures having the same class label, and perform grouping processing on the pictures with the same class label according to the similarity to obtain at least one set of pictures, where each group of pictures The similarity between any two pictures is greater than a preset threshold;

处理模块33，设置为在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像存储到云端，同时删除该组剩余的图片。The processing module 33 is configured to select a local picture that needs to be saved in the terminal device in each group of pictures, and obtain a difference image between each picture of the group and the local picture, and store the difference image to In the cloud, delete the remaining pictures of the group.

本发明实施例的图片的管理装置，还包括：The image management device of the embodiment of the present invention further includes:

本发明实施例的图片的管理装置，所述分类模块31包括：In the image management apparatus of the embodiment of the present invention, the classification module 31 includes:

划分子模块，设置为若所述检测模块检测到所述终端设备的图片中包含人脸信息，则将所述图片划分为人物类图片，若不包含人脸信息，将所述图片划分为风景类图片。a dividing sub-module, configured to: if the detecting module detects that the picture of the terminal device includes face information, divide the picture into a character-like picture, and if the face information is not included, divide the picture into a landscape Class picture.

本发明实施例的图片的管理装置，所述分组模块32包括：In the image management apparatus of the embodiment of the present invention, the grouping module 32 includes:

本发明实施例的图片的管理装置，所述提取子模块包括：In the image management apparatus of the embodiment of the present invention, the extracting submodule includes:

本发明实施例的图片的管理装置，所述获取子模块包括：In the image management apparatus of the embodiment of the present invention, the acquiring submodule includes:

第三获取单元，设置为在所述终端设备的图片中选取一张未进行分组的第一图片，并获取所述第一图片的类标签；a third acquiring unit, configured to: select a first picture that is not grouped in the picture of the terminal device, and acquire a class label of the first picture;

第一计算单元，设置为若所述类标签表明所述第一图片为人物类图片，则在所述终端设备未进行分组的图片中选取类标签为人物类图片的第二图片，并根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度；a first calculating unit, configured to: if the class label indicates that the first picture is a character class picture, select a second picture of the character class picture in the picture that is not grouped by the terminal device, and according to the picture a color histogram and a contour feature of the face region, calculating the first picture and the second Similarity of the picture;

本发明实施例的图片的管理装置，所述第一计算单元包括：In the image management apparatus of the embodiment of the present invention, the first calculating unit includes:

本发明实施例的图片的管理装置，所述处理模块33包括：In the image management apparatus of the embodiment of the present invention, the processing module 33 includes:

需要说明的是，该装置及终端是与上述方法实施例对应的装置和终端，上述方法实施例中所有实现方式均适用于该装置和终端的实施例中，也能达到相同的技术效果。It should be noted that the device and the terminal are devices and terminals corresponding to the foregoing method embodiments. All the implementation manners in the foregoing method embodiments are applicable to the device and the terminal embodiment, and the same technical effects can be achieved.

图4表示本发明实施例的终端设备的硬件结构示意图，如图4所示，该设备包括：FIG. 4 is a schematic diagram showing the hardware structure of a terminal device according to an embodiment of the present invention. As shown in FIG. 4, the device includes:

一个或多个处理器41，图4中以一个处理器41为例；One or more processors 41, one processor 41 is taken as an example in FIG. 4;

存储器42； Memory 42;

所述设备还可以包括：输入装置43和输出装置44。The apparatus may also include an input device 43 and an output device 44.

所述设备中的处理器41、存储器42、输入装置43和输出装置44可以通过总线或者其他方式连接，图4中以通过总线连接为例。The processor 41, the memory 42, the input device 43, and the output device 44 in the device may be connected by a bus or other means, as exemplified by a bus connection in FIG.

存储器42作为一种非易失性的计算机可读存储介质，可用于存储软件程序、计算机可执行程序以及模块，如本发明实施例中的相位差的确定方法对应的程序指令/模块(例如附图3所示的分类模块31、分组模块32和处理模块33)。处理器41通过运行存储在存储器42中的软件程序、指令以及模块，从而执行服务器的各种功能应用以及数据处理，即实现上述方法实施例的图片的管理方法。The memory 42 is a non-volatile computer readable storage medium, and can be used for storing a software program, a computer executable program, and a module, such as a program instruction/module corresponding to the method for determining a phase difference in the embodiment of the present invention (for example, attached) The classification module 31, the grouping module 32, and the processing module 33) shown in FIG. The processor 41 executes various functional applications and data processing of the server by running software programs, instructions, and modules stored in the memory 42, that is, the management of the picture of the above method embodiment is implemented. law.

存储器42可以包括存储程序区和存储数据区，其中，存储程序区可存储操作***、至少一个功能所需要的应用程序；存储数据区可存储根据终端设备的使用所创建的数据等。此外，存储器42可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实施例中，存储器42可选包括相对于处理器41远程设置的存储器，这些远程存储器可以通过网络连接至终端设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。The memory 42 may include a storage program area and an storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created according to usage of the terminal device, and the like. Moreover, memory 42 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, memory 42 may optionally include memory remotely located relative to processor 41, which may be connected to the terminal device over a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

输入装置43可用于接收输入的数字或字符信息，以及产生与终端的用户设置以及功能控制有关的键信号输入。输出装置44可包括显示屏等显示设备。 Input device 43 can be used to receive input numeric or character information and to generate key signal inputs related to user settings and function control of the terminal. Output device 44 may include a display device such as a display screen.

所述一个或者多个模块存储在所述存储器42中，当被所述一个或者多个处理器41执行时，执行上述实施例以及可选实时方式的图片的管理方法。The one or more modules are stored in the memory 42 and, when executed by the one or more processors 41, perform management methods of the above-described embodiments and pictures in an optional real-time manner.

本发明实施例还提供了一种非易失性的计算机可读存储介质，该非易失性的计算机存储介质存储有计算机可执行指令，该计算机可执行指令可执行上述任一实施例给出的图片的管理方法。Embodiments of the present invention also provide a non-transitory computer readable storage medium storing computer-executable instructions executable in any of the above embodiments. The management method of the picture.

本发明实施例的图片的管理方法、装置及终端设备，解决了终端设备中相似图片占用大量宝贵内存的问题，同时可以提高图片的保密度，保护用户隐私，提高用户体验。The method, device and terminal device for managing the picture in the embodiment of the invention solve the problem that a similar picture in the terminal device occupies a large amount of valuable memory, and can improve the density of the picture, protect the user privacy, and improve the user experience.

工业实用性Industrial applicability

本发明实施例的图片的管理方法，解决了终端设备中相似图片占用大量宝贵内存的问题，且本发明实施例在云端仅存储差值图像，当存储在云端的差值图像被解码后也无法看出原图像内容，有效防止因***漏洞或其他原因造成的图片泄露 The method for managing the picture in the embodiment of the present invention solves the problem that the similar picture in the terminal device occupies a large amount of valuable memory, and the embodiment of the present invention only stores the difference image in the cloud, and the difference image stored in the cloud cannot be decoded. See the original image content to effectively prevent image leaks caused by system vulnerabilities or other reasons

Claims

一种图片的管理方法，包括：A method of managing pictures, including:

对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含有表示该图片类别的类标签；Performing classification processing on the picture in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category;

获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的任意两张图片之间的相似度大于预设阈值；以及Obtaining similarity between pictures having the same class label, and grouping the pictures with the same class label according to the similarity to obtain at least one set of pictures, wherein any two pictures in each set of pictures The similarity between the two is greater than the preset threshold;

在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将所述差值图像存储到云端，同时删除该组剩余的图片。Selecting a local image to be saved in the terminal device in each group of pictures, and obtaining a difference image between each of the remaining pictures of the group and the local picture, storing the difference image in the cloud, and deleting the image The remaining pictures of the group.
根据权利要求1所述的图片的管理方法，还包括：The method for managing a picture according to claim 1, further comprising:

从云端获取所述差值图像，并将所述差值图像与对应的本地图片相加，恢复被删除的图片。The difference image is acquired from the cloud, and the difference image is added to the corresponding local picture to restore the deleted picture.
根据权利要求1所述的图片的管理方法，其中，所述对终端设备中的图片进行分类处理，得到分类后的图片，包括：The method for managing a picture according to claim 1, wherein the classifying the picture in the terminal device to obtain the classified picture comprises:

检测所述终端设备的图片中是否包含人脸信息；以及Detecting whether the face information is included in the picture of the terminal device;

若包含人脸信息，则将所述图片划分为人物类图片，若不包含人脸信息，将所述图片划分为风景类图片。If the face information is included, the picture is divided into a character type picture, and if the face information is not included, the picture is divided into a landscape picture.
根据权利要求1或3所述的图片的管理方法，其中，所述获取具有相同类标签的图片之间的相似度，包括：The method for managing a picture according to claim 1 or 3, wherein the obtaining the similarity between the pictures having the same class label comprises:

根据图片的类标签，提取终端设备中每张图片的特征信息；以及Extracting feature information of each picture in the terminal device according to the class label of the picture;

根据所述特征信息，获取具有相同类标签的图片之间的相似度。Obtaining similarities between pictures having the same class label according to the feature information.
根据权利要求4所述的图片的管理方法，其中，所述根据图片的类标签，提取每张图片的特征信息，包括：The method for managing a picture according to claim 4, wherein the extracting the feature information of each picture according to the class label of the picture comprises:

若所述图片的类标签表明所述图片为人物类图片，则提取所述图片特征信息包括图片的颜色直方图及人脸区域的轮廓特征；以及If the class tag of the picture indicates that the picture is a character class picture, extracting the picture feature information includes a color histogram of the picture and a contour feature of the face area;

若所述图片的类标签表明所述图片为风景类图片，则提取所述图片特征信息包括图片的颜色直方图。If the class label of the picture indicates that the picture is a landscape picture, extracting the picture feature information includes a color histogram of the picture.
根据权利要求4所述的图片的管理方法，其中，所述根据所述特征信息，获取具有相同类标签的图片之间的相似度，包括：The method for managing a picture according to claim 4, wherein the obtaining the similarity between the pictures having the same class label according to the feature information comprises:

在所述终端设备的图片中选取一张未进行分组的第一图片，并获取所述第一图片的类标签； Selecting a first picture that is not grouped in the picture of the terminal device, and acquiring a class label of the first picture;

若所述类标签表明所述第一图片为人物类图片，则在所述终端设备未进行分组的图片中选取类标签为人物类图片的第二图片，并根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度；以及If the class label indicates that the first picture is a character class picture, selecting a second picture of the character class picture in the picture that is not grouped by the terminal device, and according to the color histogram and the face of the picture a contour feature of the region, calculating a similarity between the first picture and the second picture;

若所述类标签表明所述第一图片为风景类图片，则在所述终端设备未进行分组的图片中选取类标签为风景类图片的第三图片，并根据图片的颜色直方图，计算所述第三图片与所述第一图片的相似度。If the class label indicates that the first picture is a landscape picture, select a third picture whose class label is a landscape picture in a picture that is not grouped by the terminal device, and calculate a location according to a color histogram of the picture. The similarity between the third picture and the first picture is described.
根据权利要求6所述的图片的管理方法，其中，所述根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度，包括：The method for managing a picture according to claim 6, wherein the calculating the similarity between the first picture and the second picture according to the color histogram of the picture and the contour feature of the face area comprises:

判断所述第一图片与所述第二图片中的人脸个数是否相同；以及Determining whether the number of faces in the first picture and the second picture are the same;

若相同，则通过公式S＝r1*P+r2*H计算所述第一图片和所述第二图片的相似度；If the same, the similarity between the first picture and the second picture is calculated by the formula S=r1*P+r2*H;

其中，S表示相似度，r1和r2为预设系数，且r1+r2＝1，P表示人脸区域的轮廓特征的匹配度，H表示颜色直方图相似度。Where S represents the similarity, r1 and r2 are preset coefficients, and r1+r2=1, P represents the matching degree of the contour feature of the face region, and H represents the color histogram similarity.
根据权利要求1所述的图片的管理方法，其中，所述在每组图片中选取出需要保存在所述终端设备中的本地图片，包括：The method for managing a picture according to claim 1, wherein the selecting a local picture to be saved in the terminal device in each group of pictures comprises:

获取每组图片的图片质量；以及Get the image quality of each set of images; and

在每组图片中，选取图片质量最高的一张图片作为所述本地图片。In each set of pictures, one picture with the highest picture quality is selected as the local picture.
一种图片的管理装置，包括：A picture management device comprising:

分类模块，设置为对终端设备中的图片进行分类处理，得到分类后的图片，所述分类后的图片包含表示该图片类别的类标签；a classification module, configured to classify a picture in the terminal device to obtain a classified picture, where the classified picture includes a class label indicating the picture category;

分组模块，设置为获取具有相同类标签的图片之间的相似度，并根据所述相似度对所述具有相同类标签的图片进行分组处理，得到至少一组图片，其中，每组图片中的两张图片之间的相似度大于预设阈值；以及a grouping module, configured to acquire a similarity between pictures having the same class label, and perform grouping processing on the pictures with the same class label according to the similarity to obtain at least one group of pictures, where each group of pictures The similarity between the two images is greater than a preset threshold;

处理模块，设置为在每组图片中选取出需要保存在所述终端设备中的本地图片，并获取该组剩余的每张图片与所述本地图片的差值图像，将差值图像存储到云端，同时删除该组剩余的图片。a processing module, configured to select a local picture that needs to be saved in the terminal device in each group of pictures, and obtain a difference image between each picture of the group and the local picture, and store the difference image in the cloud , while deleting the remaining pictures of the group.
根据权利要求9所述的图片的管理装置，还包括：The picture management device according to claim 9, further comprising:

恢复模块，设置为从云端获取所述差值图像，并将所述差值图像与对应的本地图片相加，恢复被删除的图片。And a recovery module, configured to acquire the difference image from the cloud, and add the difference image to the corresponding local image to recover the deleted picture.
根据权利要求9所述的图片的管理装置，其中，所述分类模块包括： The picture management device according to claim 9, wherein the classification module comprises:

检测子模块，设置为检测所述终端设备的图片中是否包含人脸信息；以及a detecting submodule configured to detect whether the face information of the terminal device includes a face information;

划分子模块，设置为若所述检测模块检测到所述终端设备的图片中包含人脸信息，则将所述图片划分为人物类图片，若不包含人脸信息，将所述图片划分为风景类图片。a dividing sub-module, configured to: if the detecting module detects that the picture of the terminal device includes face information, divide the picture into a character-like picture, and if the face information is not included, divide the picture into a landscape Class picture.
根据权利要求11所述的图片的管理方法，其中，所述分组模块包括：The method of managing a picture according to claim 11, wherein the grouping module comprises:

提取子模块，设置为根据图片的类标签，提取终端设备中每张图片的特征信息；以及Extracting a sub-module, configured to extract feature information of each picture in the terminal device according to the class label of the picture;

获取子模块，设置为根据所述特征信息，获取具有相同类标签的图片之间的相似度。The obtaining submodule is configured to acquire the similarity between the pictures having the same class label according to the feature information.
根据权利要求12所述的图片的管理装置，其中，所述提取子模块包括：The picture management device according to claim 12, wherein the extraction submodule comprises:

第一获取单元，设置为若所述图片的类标签表明所述图片为人物类图片，则获取所述图片的颜色直方图及人脸区域的轮廓特征；以及a first acquiring unit, configured to acquire a color histogram of the picture and a contour feature of the face area if the class label of the picture indicates that the picture is a character class picture;

第二获取单元，设置为若所述图片的类标签表明所述图片为风景类图片，则获取所述图片的颜色直方图。The second obtaining unit is configured to obtain a color histogram of the picture if the class label of the picture indicates that the picture is a landscape picture.
根据权利要求12所述的图片的管理装置，其中，所述获取子模块包括：The picture management apparatus according to claim 12, wherein the obtaining submodule comprises:

第三获取单元，设置为在所述终端设备的图片中选取一张未进行分组的第一图片，并获取所述第一图片的类标签；以及a third acquiring unit, configured to: select a first picture that is not grouped in the picture of the terminal device, and acquire a class label of the first picture;

第一计算单元，设置为若所述类标签表明所述第一图片为人物类图片，则在所述终端设备未进行分组的图片中选取类标签为人物类图片的第二图片，并根据图片的颜色直方图及人脸区域的轮廓特征，计算所述第一图片和所述第二图片的相似度；a first calculating unit, configured to: if the class label indicates that the first picture is a character class picture, select a second picture of the character class picture in the picture that is not grouped by the terminal device, and according to the picture a color histogram and a contour feature of the face region, and calculating a similarity between the first picture and the second picture;

第二计算单元，设置为若所述类标签表明所述第一图片为风景类图片，则在所述终端设备未进行分组的图片中选取类标签为风景类图片的第三图片，并根据图片的颜色直方图，计算所述第三图片与所述第一图片的相似度。a second calculating unit, configured to: if the class label indicates that the first picture is a landscape picture, select a third picture whose class label is a landscape picture in a picture that is not grouped by the terminal device, and according to the picture a color histogram, calculating a similarity between the third picture and the first picture.
根据权利要求14所述的图片的管理装置，其中，所述第一计算单元包括：The picture management device according to claim 14, wherein the first calculation unit comprises:

判断子单元，设置为判断所述第一图片与所述第二图片中的人脸个数是否相同；以及a determining subunit, configured to determine whether the number of faces in the first picture and the second picture are the same;

计算子单元，设置为若所述第一图片与所述第二图片中的人脸个数相同，则通过公式S＝r1*P+r2*H计算所述第一图片和所述第二图片的相似度；Calculating a subunit, configured to calculate the first picture and the second picture by using a formula S=r1*P+r2*H if the number of faces in the first picture and the second picture are the same Similarity

其中，S表示相似度，r1和r2为预设系数，且r1+r2＝1，P表示人脸区域的轮廓特征的匹配度，H表示颜色直方图相似度。Where S represents the similarity, r1 and r2 are preset coefficients, and r1+r2=1, P represents the face region The degree of matching of the contour features, H represents the color histogram similarity.
根据权利要求9所述的图片的管理装置，其中，所述处理模块包括：The picture management device according to claim 9, wherein the processing module comprises:

第四获取子模块，设置为获取每组图片的图片质量；以及a fourth obtaining submodule, configured to obtain a picture quality of each group of pictures;

选取子模块，设置为在每组图片中，选取图片质量最高的一张图片作为所述本地图片。The sub-module is selected, and in each set of pictures, one picture with the highest picture quality is selected as the local picture.
一种终端设备，包括如权利要求9-16任一项所述的图片的管理装置。A terminal device comprising the picture management device according to any one of claims 9-16.
一种非易失性计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如权利要求1-12任一项所述的图片的管理方法。 A non-transitory computer readable storage medium storing computer executable instructions for performing the method of managing a picture according to any of claims 1-12.