CN108712655B - group image coding method for similar image set merging - Google Patents
group image coding method for similar image set merging Download PDFInfo
- Publication number
- CN108712655B CN108712655B CN201810507677.2A CN201810507677A CN108712655B CN 108712655 B CN108712655 B CN 108712655B CN 201810507677 A CN201810507677 A CN 201810507677A CN 108712655 B CN108712655 B CN 108712655B
- Authority
- CN
- China
- Prior art keywords
- image
- image set
- node
- priority
- root node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/154—Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/182—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a pixel
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/96—Tree coding, e.g. quad-tree coding
Abstract
The invention provides group image coding methods for similar image set merging, which comprises the steps of selecting a main image set, determining merging priorities of the other image sets, merging the other image sets with the main image set in sequence from high to low in priority, performing branch search on the tree graph of the main image set layer by layer when the image sets with each priority are merged, finding a parent node in the tree graph of the main image set for a root node of the image set with the priority, and coding images corresponding to the root nodes of non-main image sets.
Description
Technical Field
The invention belongs to the technical field of image coding, and particularly relates to group image coding methods for similar image set merging, which can be used for merging similar image sets in scenes of cloud group image management and image database management.
Background
In recent decades, with the rapid development of the internet-related industry, multimedia contents such as digital images have also been increasing explosively, according to Facebook report of the world's largest social network service company, stored photos have been increased by over two hundred billion and are increasing by three hundred billion per day.
Group image compression coding is mainly applied to image sets with correlation at present, so that group image compression coding research is mainly focused on two directions:
the current state of the research of this kind of problem is that Yurij S.Musatenko proposes that KL (Karhunen-Loeve) transform is used to realize the best Image coding (Optimal Image coding using Karhunen-Loeve transform, OIL) so that the maximum average contribution of each Image can be obtained by the odd function of KL, and the contribution is the fastest in all basis functions[i](set redundancy compression, SRC) reduces redundancy between images.
The method in the above is easy to understand, and has good effect when applied to the medical and satellite image field, but in real life, the scene contained in the image is not as single as as the image in the medical and satellite field, but changes greatly, so it is difficult to find representative images as reference images of all the rest images in the image set, so that the effect of encoding and compressing the image set is not improved greatly.
The second research direction is to store images in a personal album image set and cloud images, wherein identical scenes are usually shot from different angles or under different brightness, the similarity between the images is high, each image set is compressed and coded by utilizing the correlation between the internal images of the image set, at present, the images in many application scenes belong to the same type, such as the field of image retrieval and image recognition, group image coding is to code by utilizing the similarity between the images in the image set, firstly, the similarity between the images is quantitatively described by utilizing an algorithm of related image processing, a corresponding reference coding structure is generated according to the similarity between the images, a reference picture of each image to be coded, namely a target coding image, in the image set to be coded is determined, then, the step is carried out to carry out geometric deformation and luminosity transformation to enable the transformed images to be closer to the target images, and finally, the images in the image set are coded by adopting a video coding technology.
Similar image sets may exist in the group images, and if the images are combined, more storage space can be saved; the image sets added to the image library may also be image sets subjected to group image coding, so that the similar image sets are combined, and the coding efficiency can be improved better.
For example, the method comprises methods for compressing and encoding Cloud Storage image sets based on local features, which are disclosed in a paper "Photo Album Compression for Cloud Storage Using LocalFeatures" published in 2014 by Zhongbo Shi, Xiaoyan Sun, and andFeng Wu in IEEE Journal on generating and Selected Topica in Circuits and systems.
Disclosure of Invention
The invention aims to provide a group image coding method for merging similar image sets aiming at the defects of the prior art and improve the coding efficiency.
In order to achieve the purpose, the technical scheme adopted by the invention comprises the following steps:
(1) initializing limiting parameters:
giving a value of the limiting depth d;
(2) determining a main image set of an image set to be merged:
(2a) reading the height of each image set tree graph in the image sets to be combined, and selecting the image set corresponding to the tree graph with the highest height;
(2b) if the number of the image sets corresponding to the highest tree graph is , the image set is taken as the main image set of the image set to be merged, if the number of the image sets corresponding to the highest tree graph is multiple, and when the number of the image sets containing the largest number of images is , the image set is taken as the main image set of the image set to be merged, otherwise, when the number of the image sets containing the largest number of images is multiple, the images corresponding to the root nodes of the tree graphs of the image sets are decoded, the distances among the images corresponding to the root nodes are calculated, the sum of the distances from each root node to the rest root nodes is solved, and the image set corresponding to the root node with the minimum sum of the distances from the rest root nodes is selected as the main image set;
(3) determining the merging priority of the image sets except the main image set in the image set to be merged:
determining the merging priority of each image set except the main image set in the image set to be merged; recording the heights of the tree graphs of the image sets as h1 and h2 … … hn in sequence according to the priority of the image sets;
(4) setting iteration parameters:
setting the depth of an initial node as i, and setting i as 1; the priority of the image set merged with the main image set at each time is j, and j is made to be 2;
(5) calculating the distance from the initial node to the root node of the image set with the priority of j:
(5a) taking a root node of the main image set tree diagram as an initial node, and decoding the root node to obtain a decoded image of the initial node;
(5b) calculating the distance from the image corresponding to the initial node to the image corresponding to the root node Rj of the image set tree graph with the priority of j through the decoded image of the initial node, and taking the calculation result as the distance from the initial node to the root node of the image set with the priority of j;
(6) calculating the distance from the child node of the initial node to the root node of the image set with the priority of j:
(6a) decoding images corresponding to child nodes of the initial node to obtain decoded images of the child nodes;
(6b) calculating the distance from the image corresponding to the child node of the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the child node, and taking the calculation result as the distance from the child node of the initial node to the root node of the image set with the priority of j;
(7) judging whether the node with the minimum distance to the root node Rj is an initial node or not:
judging whether the minimum distance from the initial node and child nodes thereof to the root node Rj is the distance from the initial node to the root node Rj, if so, taking the initial node as a father node of the Rj, connecting the root node Rj to the initial node, and obtaining a tree diagram formed by combining an image set with the priority j and a main image set; otherwise, taking the node with the minimum distance to the root node Rj of the tree graph with the priority of the j image set as an initial node, and executing the step (8);
(8) judging whether the depth i of the initial node is less than d-hj-1:
if so, executing the step (6), otherwise, taking the initial node as a parent node of a root node Rj, connecting the root node Rj to the initial node to obtain a tree graph formed by merging an image set with the priority j and a main image set, and executing the step (9);
(9) judging whether j is equal to the number of the image sets to be merged:
judging whether j is equal to the number of the image sets to be merged, if so, merging all the image sets to be merged, taking the tree graph obtained by merging the image set with the priority of j and the main image set as a final tree graph, and executing the step (10); otherwise, let j equal to j +1, and take the image set obtained by combining the image set with the priority j and the main image set as the main image set, and execute step (5);
(10) encoding an image corresponding to a root node of a non-main image set:
(10a) determining a reference image of an image corresponding to a root node of a non-main image set according to the obtained tree diagram;
(10b) calculating a perspective transformation matrix H between the reference image and the image corresponding to the non-main image set root node;
(10c) performing geometric transformation on the reference image according to the perspective transformation matrix;
(10d) performing photometric transformation on the reference image after the geometric transformation;
(10e) and performing inter-frame coding on the image corresponding to the root node of the non-main image set by using the reference image after luminosity transformation.
Compared with the prior art, the invention has the following advantages:
when the similar image sets are coded, the similar image sets are combined by utilizing the relation among the similar image sets, the images corresponding to the root nodes of the non-main image sets are subjected to inter-frame coding, and compared with the method that the similar image sets are respectively coded and the images corresponding to the root nodes of the non-main image sets are subjected to intra-frame coding in the prior art, the coding efficiency of the group images is improved.
Drawings
FIG. 1 is a block diagram of an implementation flow of the present invention;
fig. 2 to 5 are graphs comparing the encoding efficiency of the image sets mainbuilding, defense, road (more), and road (less) after merging with that before merging according to the present invention.
Detailed Description
The invention will now be described in further detail with reference to the figures and specific embodiments.
referring to fig. 1:
step 1) initializing limiting parameters:
giving a value of the limiting depth d;
step 2) determining a main image set of an image set to be merged:
step 2a) reading the height of each image set treemap in the image sets to be combined, and selecting the image set corresponding to the treemap with the highest height;
step 2b) if the image sets corresponding to the highest tree graphs are , taking the image sets as main image sets of image sets to be merged, if the image sets corresponding to the highest tree graphs are multiple, and when the image sets containing the largest number of images are , taking the image sets as the main image sets of the image sets to be merged, otherwise, when the image sets containing the largest number of images are multiple, decoding the images corresponding to the root nodes of the tree graphs of the image sets, calculating the distances among the images corresponding to the root nodes, solving the sum of the distances from each root node to the rest root nodes, and selecting the image set corresponding to the root node with the minimum sum of the distances from the rest root nodes as the main image set;
the distance between the images corresponding to the root nodes is calculated by adopting the SIFT distance, and the calculation formula is as follows:
wherein D isiThe distance between any pairs of SIFT feature matching points between images corresponding to two nodes is defined, and n is the number of pairs of SIFT feature matching points between images corresponding to any two nodes.
Step 3) determining the merging priority of the image sets except the main image set in the image set to be merged:
step 3a), firstly, determining the priority of each image set according to the height of the tree-shaped graph of the image sets, wherein the image set with the larger tree-shaped graph height has the priority higher than the image set with the smaller tree-shaped graph height;
step 3b), when image sets with the same height of the tree-shaped graph exist, determining the priority of the image sets according to the number of images contained in the image sets, wherein the image sets with more images have higher priority than the image sets with less images; if image sets with the same number of images still exist, determining the priority of the image sets with the same number of images according to the sum of the distances from the root node of the tree-shaped graph to the root node of the tree-shaped graph of the image set with the priority not determined, wherein the priority with smaller sum of the distances from the root node of the tree-shaped graph to the root node of the tree-shaped graph of the image set with the priority not determined is higher;
because the image in the image set merged first may be used as a reference image of the image corresponding to the root node of the tree graph of the image set with lower priority in the subsequent merging process. Therefore, the priority of image set combination is determined according to the rule, so that the cost of the tree graph of the combined image set is lower, and the coding efficiency of the combined image set is higher. The priority of the other image sets is sequentially increased on the basis of the main image set; recording the heights of the tree graphs of the image sets as h1 and h2 … … hn in sequence according to the priority of the image sets;
step 4), setting iteration parameters:
setting the depth of an initial node as i, and setting i as 1; the priority of the image set merged with the main image set at each time is j, and j is made to be 2; since the priority of the main image set is 1, merging is started from the image set with the priority of 2;
step 5) calculating the distance from the initial node to the image set root node with the priority of j:
step 5a) taking a root node of the main image set tree diagram as an initial node, and decoding the root node to obtain a decoded image of the initial node;
step 5b) calculating the distance from the image corresponding to the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the initial node, and taking the calculation result as the distance from the initial node to the root node of the image set with the priority of j, wherein the distance is the SIFT distance;
step 6) calculating the distance from the child node of the initial node to the root node of the image set with the priority of j:
step 6a) decoding the image corresponding to the child node of the initial node to obtain a decoded image of the child node;
step 6b) calculating the distance from the image corresponding to the child node of the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the child node, and taking the calculation result as the distance from the child node of the initial node to the root node of the image set with the priority of j;
step 7) judging whether the node with the minimum distance to the root node Rj is an initial node:
judging whether the minimum distance from the initial node and child nodes thereof to the root node Rj is the distance from the initial node to the root node Rj, if so, taking the initial node as a father node of the Rj, connecting the root node Rj to the initial node, and obtaining a tree diagram formed by combining an image set with the priority j and a main image set; because the depth of the initial node is smaller than that of the child nodes of the initial node, and the distance from the initial node to the root node Rj is smaller, the image corresponding to the initial node can be directly selected as the reference image of the root node Rj, and the complexity of the algorithm is reduced; otherwise, taking the node with the minimum distance to the root node Rj of the tree graph with the priority of the j image set as an initial node, and performing the step (8);
step 8) judging whether the depth i of the initial node is less than d-hj-1:
if so, executing the step (6), otherwise, in order to meet the depth limitation, taking the initial node as a parent node of a root node Rj, connecting the root node Rj to the initial node, obtaining a tree graph formed by combining an image set with the priority j and a main image set, and executing the step (9);
step 9) judging whether j is equal to the number of image sets to be merged, if so, merging all image sets to be merged, taking a tree graph obtained by merging the image set with the priority of j and the main image set as a final tree graph, and executing the step 10, otherwise, enabling j to be j +1, calculating the distance from the image decoded before to the image corresponding to the root node Rj of the tree graph with the priority of j, simultaneously taking the image set obtained by merging the image set with the priority of j and the main image set as the main image set, and executing the step 5, wherein in the cyclic process of merging each priority image set, if the image decoded before does not need to be decoded, the calculated image distance does not need to be repeatedly calculated, and even if the image set is not the child node of the initial node at the depth of , the calculated distance is participated in the selection of the minimum distance;
step 10) encoding the image corresponding to the root node of the non-main image set:
step 10a) determining a reference image of an image corresponding to a root node of a non-main image set according to the obtained attribute map;
step 10b) calculating a perspective transformation matrix H between the reference image and the image corresponding to the non-main image set root node;
step 10c) geometrically transforming the reference image according to the perspective transformation matrix;
step 10d) performing luminosity transformation on the reference image after the geometric transformation;
and step 10e) utilizing the reference image after luminosity transformation to perform interframe coding on the image corresponding to the root node of the non-main image set.
Referring to fig. 1:
step 1) initializing limiting parameters:
giving a value of the limiting depth d;
step 2) determining a main image set of an image set to be merged:
step 2a) because only two image sets exist in the embodiment, when the two image sets are different in height, the image set corresponding to the tree graph with the highest height is selected as the main image set of the image set to be merged;
step 2b), when the heights of the two image sets are the same, selecting the image set containing more images as a main image set of the image set to be merged, if the numbers of the images contained in the two image sets are still the same, decoding the images corresponding to the root nodes of the tree graphs of the two image sets, calculating the distance between the images corresponding to the root nodes, selecting the image set corresponding to the root node with smaller distance to another node as a main image set, and setting the priority of the main image set to be 1;
the distance between the images corresponding to the root nodes is calculated by adopting the SIFT distance, and the calculation formula is as follows:
wherein D isiIs the distance between any pairs of SIFT feature matching points between the images corresponding to the two nodes, and n is the SIFT feature between the images corresponding to the two nodesAnd (5) figuring the number of the matching point pairs.
Step 3) determining the merging priority of the rest image sets:
because only two image sets need to be merged, the priority of the main image set is 1, and the priority of the other image set is 2, and the heights of the image set tree graphs are sequentially recorded as h1 and h2 according to the priority of the image sets;
step 4), setting iteration parameters:
setting the depth of an initial node as i, and setting i as 1; the priority of the image set merged with the main image set at each time is j, and j is made to be 2; since the priority of the main image set is 1, merging is started from the image set with the priority of 2;
step 5) calculating the distance from the initial node to the image set root node with the priority of j:
step 5a) taking a root node of the main image set tree diagram as an initial node, and decoding the root node to obtain a decoded image of the initial node;
step 5b) calculating the distance from the image corresponding to the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the initial node, and taking the calculation result as the distance from the initial node to the root node of the image set with the priority of j, wherein the distance is the SIFT distance;
step 6) calculating the distance from the child node of the initial node to the root node of the image set with the priority of j:
step 6a) decoding the image corresponding to the child node of the initial node to obtain a decoded image of the child node;
step 6b) calculating the distance from the image corresponding to the child node of the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the child node, and taking the calculation result as the distance from the child node of the initial node to the root node of the image set with the priority of j;
step 7) judging whether the node with the minimum distance to the root node Rj is an initial node:
judging whether the minimum distance from the initial node and the child nodes thereof to the root node Rj is the distance from the initial node to the root node Rj, if so, taking the initial node as a father node of the Rj, connecting the root node Rj to the initial node, and obtaining a tree-shaped graph formed by combining an image set with the priority of j and a main image set, wherein the depth of the initial node is smaller than that of the child nodes thereof, and the distance from the initial node to the root node Rj is smaller, so that an image corresponding to the initial node can be directly selected as a reference image of the root node Rj, and the complexity of the algorithm is reduced; otherwise, taking the node with the minimum distance to the root node Rj of the tree graph with the priority of the j image set as an initial node, and performing the step (8);
step 8) judging whether the depth i of the initial node is less than d-hj-1:
if so, executing the step (6), otherwise, in order to meet the depth limitation, taking the initial node as a parent node of a root node Rj, connecting the root node Rj to the initial node, and obtaining a tree graph formed by combining an image set with the priority of j and a main image set; since the embodiment only merges two similar image sets, a tree graph after the image sets are merged is obtained.
Step 9) encoding the image corresponding to the root node of the non-main image set:
step 9a) determining a reference image of an image corresponding to a root node of a non-main image set according to the obtained tree diagram;
step 9b) calculating a perspective transformation matrix H between the reference image and the image corresponding to the non-main image set root node;
step 9c) performing geometric transformation on the reference image according to the perspective transformation matrix;
step 9d) performing luminosity transformation on the reference image after the geometric transformation;
and 9e) utilizing the reference image after luminosity transformation to perform interframe coding on the image corresponding to the root node of the non-main image set.
The technical effect of the invention is further illustrated in steps through simulation experiments as follows:
1) the experimental conditions are as follows:
the experiment is carried out in the environment of a windows Server 2008 system, a processor Intel (R) Xeon (R), a CPU E5-2650v2@2.60GHz and a RAM 64 GB. The programming language is C + +, and the programming software is VS 2010.
The details of the image sets to be combined are shown in table 1:
TABLE 1
In the above 4 test cases, rod (less) and rod (more) are the same image sets, and partial images are selected from them to form contrast rod (less).
2) And (3) analyzing the experimental content and the result:
the invention and the prior art are respectively adopted to code the test cases to obtain an image coding efficiency comparison graph and a non-main image set root node coding efficiency comparison graph in the prior art. As shown in fig. 2(a) to 5 (b). Fig. 2(a), fig. 3(a), fig. 4(a), and fig. 5(a) are respectively a comparison graph of coding efficiency of all images formed by merging a mainbuilding image set, a defenses image set, a road (less) image set, and a road (more) image set according to the present invention with a prior art, and fig. 2(b), fig. 3(b), fig. 4(b), and fig. 5(b) are comparison graphs of coding efficiency of images corresponding to non-main image set root nodes formed by merging a mainbuilding image set, a defenses image set, a road (less) image set, and a road (more) image set according to the present invention with a prior art. In fig. 2(a) -5 (b), the higher the peak snr is, the higher the coding efficiency is at the same bit rate.
As shown in fig. 2(a) to fig. 5(a), the difference between the coding efficiency of the image in the present invention and that of the prior art is small, because the present invention and the prior art only have a coding efficiency difference between images corresponding to the root node of the tree-shaped graph of the non-master image set, and because the number of images in the image set is large, the difference cannot be reflected when the coding efficiency of all images in the image is compared as a whole. Therefore, fig. 2(b) to fig. 5(b) compare the coding efficiency of the images corresponding to the root node of the non-main image set, and it can be seen that the coding efficiency is improved by the present invention.
In conclusion, the invention can improve the coding efficiency of the image under the condition of low complexity. The method can be used for similar image set combination in similar scenes of cloud group image management and image database management.
Claims (1)
1, A group image coding method for similar image set merging, comprising the following steps:
(1) initializing limiting parameters:
giving a value of the limiting depth d;
(2) determining a main image set of an image set to be merged:
(2a) reading the height of each image set tree graph in the image sets to be combined, and selecting the image set corresponding to the tree graph with the highest height;
(2b) if the number of the image sets corresponding to the highest tree graph is , the image set is taken as a main image set of image sets to be merged, if the number of the image sets corresponding to the highest tree graph is multiple, and when the number of the image sets containing the largest number of images is , the image set is taken as the main image set of the image sets to be merged, otherwise, when the number of the image sets containing the largest number of images is multiple, the images corresponding to the root nodes of the tree graphs of the image sets are decoded, the distances among the images corresponding to the root nodes are calculated, the sum of the distances from each root node to the rest root nodes is calculated, the image set corresponding to the root node with the minimum sum of the distances to the rest root nodes is selected as a main image set, the priority of the main image set is set 1, wherein the distances among the images corresponding to the root nodes are calculated, and the calculation formula is as follows:
wherein D isiThe distance between any pairs of SIFT feature matching points between the images corresponding to the two nodes is shown, and n is the number of SIFT feature matching point pairs between the images corresponding to the two nodes;
(3) determining the merging priority of the image sets except the main image set in the image set to be merged:
determining the merging priority of each image set except the main image set in the image set to be merged; recording the heights of the tree graphs of the image sets as h1 and h2 … … hn in sequence according to the priority of the image sets;
(4) setting iteration parameters:
setting the depth of an initial node as i, and setting i as 1; the priority of the image set merged with the main image set at each time is j, and j is made to be 2;
(5) calculating the distance from the initial node to the root node of the image set with the priority of j:
(5a) taking a root node of the main image set tree diagram as an initial node, and decoding the root node to obtain a decoded image of the initial node;
(5b) calculating the distance from the image corresponding to the initial node to the image corresponding to the root node Rj of the image set tree graph with the priority of j through the decoded image of the initial node, and taking the calculation result as the distance from the initial node to the root node of the image set with the priority of j;
(6) calculating the distance from the child node of the initial node to the root node of the image set with the priority of j:
(6a) decoding images corresponding to child nodes of the initial node to obtain decoded images of the child nodes;
(6b) calculating the distance from the image corresponding to the child node of the initial node to the image corresponding to the root node Rj of the image set tree diagram with the priority of j through the decoded image of the child node, and taking the calculation result as the distance from the child node of the initial node to the root node of the image set with the priority of j;
(7) judging whether the node with the minimum distance to the root node Rj is an initial node or not:
judging whether the minimum distance from the initial node and child nodes thereof to the root node Rj is the distance from the initial node to the root node Rj, if so, taking the initial node as a father node of the Rj, connecting the root node Rj to the initial node, and obtaining a tree diagram formed by combining an image set with the priority j and a main image set; otherwise, taking the node with the minimum distance to the root node Rj of the tree graph with the priority of the j image set as an initial node, and executing the step (8);
(8) judging whether the depth i of the initial node is less than d-hj-1:
if so, executing the step (6), otherwise, taking the initial node as a parent node of a root node Rj, connecting the root node Rj to the initial node to obtain a tree graph formed by merging an image set with the priority j and a main image set, and executing the step (9);
(9) judging whether j is equal to the number of the image sets to be merged:
judging whether j is equal to the number of the image sets to be merged, if so, merging all the image sets to be merged, taking the tree graph obtained by merging the image set with the priority of j and the main image set as a final tree graph, and executing the step (10); otherwise, let j equal to j +1, and take the image set obtained by combining the image set with the priority j and the main image set as the main image set, and execute step (5);
(10) encoding an image corresponding to a root node of a non-main image set:
(10a) determining a reference image of an image corresponding to a root node of a non-main image set according to the obtained tree diagram;
(10b) calculating a perspective transformation matrix H between the reference image and the image corresponding to the non-main image set root node;
(10c) performing geometric transformation on the reference image according to the perspective transformation matrix;
(10d) performing photometric transformation on the reference image after the geometric transformation;
(10e) and performing inter-frame coding on the image corresponding to the root node of the non-main image set by using the reference image after luminosity transformation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810507677.2A CN108712655B (en) | 2018-05-24 | 2018-05-24 | group image coding method for similar image set merging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810507677.2A CN108712655B (en) | 2018-05-24 | 2018-05-24 | group image coding method for similar image set merging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108712655A CN108712655A (en) | 2018-10-26 |
| CN108712655B true CN108712655B (en) | 2020-01-31 |
Family
ID=63869476
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810507677.2A Active CN108712655B (en) | 2018-05-24 | 2018-05-24 | group image coding method for similar image set merging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108712655B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111416976B (en) * | 2019-01-08 | 2023-12-08 | 浙江大学 | Video decoding method, video encoding method, device, equipment and storage medium |
| CN115604480B (en) * | 2022-09-19 | 2023-06-20 | 杨凌职业技术学院 | Cloud image set encoding method for multi-priority similar images |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060104484A1 (en) * | 2004-11-16 | 2006-05-18 | Bolle Rudolf M | Fingerprint biometric machine representations based on triangles |
| CN106028045B (en) * | 2010-04-13 | 2019-06-25 | Ge视频压缩有限责任公司 | The method of decoding data stream, the method and its decoder for generating data flow |
| US9367939B2 (en) * | 2013-10-22 | 2016-06-14 | Nokia Technologies Oy | Relevance based visual media item modification |
| CN106503734B (en) * | 2016-10-14 | 2019-08-06 | 哈尔滨工程大学 | Image classification method based on trilateral filter and the sparse autocoder of storehouse |
| CN107105306B (en) * | 2017-05-16 | 2019-10-11 | 西安电子科技大学 | Group's picture coding structure generation method based on Zhu Liu method and OTTC |
-
2018
- 2018-05-24 CN CN201810507677.2A patent/CN108712655B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108712655A (en) | 2018-10-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Blueprint separable residual network for efficient image super-resolution | |
| Lin et al. | Gait recognition with multiple-temporal-scale 3d convolutional neural network | |
| CN109257604B (en) | Color attribute coding method based on TMC3 point cloud encoder | |
| JP6333190B2 (en) | How to query a video database | |
| US9054876B1 (en) | Fast efficient vocabulary computation with hashed vocabularies applying hash functions to cluster centroids that determines most frequently used cluster centroid IDs | |
| Duan et al. | Compact descriptors for visual search | |
| CN110162665B (en) | Video searching method, computer device and storage medium | |
| CN109783691B (en) | Video retrieval method for deep learning and Hash coding | |
| Zhang et al. | A joint compression scheme of video feature descriptors and visual content | |
| CN110163061B (en) | Method, apparatus, device and computer readable medium for extracting video fingerprint | |
| CN107180079B (en) | Image retrieval method based on convolutional neural network and tree and hash combined index | |
| CN108712655B (en) | group image coding method for similar image set merging | |
| CN111246287A (en) | Video processing method, video publishing method, video pushing method and devices thereof | |
| CN106997381B (en) | Method and device for recommending movies to target user | |
| CN111382602A (en) | Cross-domain face recognition algorithm, storage medium and processor | |
| Tang et al. | Deep unsupervised key frame extraction for efficient video classification | |
| CN111914938A (en) | Image attribute classification and identification method based on full convolution two-branch network | |
| CN108764258B (en) | Optimal image set selection method for group image insertion | |
| Perra et al. | Cloud-scale Image Compression Through Content Deduplication. | |
| CN113590898A (en) | Data retrieval method and device, electronic equipment, storage medium and computer product | |
| CN101464909A (en) | Fast robust approximately same video detection and exclusion method | |
| CN106611043B (en) | Video searching method and system | |
| CN107194961B (en) | Method for determining multiple reference images in group image coding | |
| US10218987B2 (en) | Methods, systems, and computer readable media for performing image compression | |
| CN115272768A (en) | Content identification method, device, equipment, storage medium and computer program product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |