CN-121985136-A - Video compression and intelligent reconstruction method and system for extremely low bandwidth

Abstract

The invention discloses a video compression and intelligent reconstruction method and system for extremely low bandwidth, which relate to the technical field of video processing and comprise the steps of collecting an original video frame sequence, constructing a session time window, executing semantic analysis of video coding on video contents in the session time window to generate a semantic recoverable grade diagram, dividing the video contents into video areas with different recoverable grades according to the semantic recoverable grade diagram, carrying out hierarchical organization on the video contents to generate bottom-protecting content data and enhanced description data, executing content self-adaptive code rate allocation on the bottom-protecting content data and the enhanced description data under the constraint of extremely low bandwidth, and respectively carrying out compression coding to form a bottom-protecting code stream and an enhanced code stream. The invention can preferentially ensure the stable reconstruction of the key semantic region by constructing the semantic recoverable grade and carrying out grading treatment on the video content and the limited bandwidth resource.

Inventors

• LI WEN
• Si Gangjun

Assignees

• 北京数智星通科技有限公司

Dates

Publication Date

20260505

Application Date

20260126

Claims (10)

1. 1. The video compression and intelligent reconstruction method for extremely low bandwidth is characterized by comprising the following steps of, Collecting an original video frame sequence, constructing a session time window, performing semantic analysis of video coding on video content in the session time window, and generating a semantic recoverable grade diagram; Dividing video content into video areas with different recoverable grades according to the semantic recoverable grade diagram, and carrying out hierarchical organization on the video content to generate bottom-protecting content data and enhancement description data; under the constraint of a target extremely low bandwidth, performing content self-adaptive code rate allocation on the bottom-protecting content data and the enhancement description data and respectively performing compression coding to form a bottom-protecting code stream and an enhancement code stream; Receiving and decoding the bottom-protected code stream and the enhanced code stream at a decoding end, and simultaneously executing intelligent reconstruction processing according to the corresponding recoverable grade constraint by combining semantic guide information carried in the enhanced code stream to generate a target quality video sequence; And performing task completion degree evaluation based on the target quality video sequence, and adaptively updating semantic recoverable grade division rules and code rate allocation strategies in a session time window.
2. 2. The method for very low bandwidth video compression and intelligent reconstruction as set forth in claim 1, wherein said sequence of original video frames comprises a plurality of frames of video image data and time identification information.
3. 3. The method for video compression and intelligent reconstruction with very low bandwidth as in claim 2, wherein the constructing the session time window comprises the following steps, Sequentially traversing the original video frame sequence according to the time identification information, and calculating brightness change, chromaticity change and spatial structure change between adjacent video frames frame by frame to form a content change measure; acquiring a change judgment threshold value based on the statistical distribution characteristic of the content change measurement, judging the position of the content change measurement exceeding the change judgment threshold value as a session boundary, and dividing the session boundary into an initial session time window; And counting the time distribution density and content change continuity of the video frames in the window aiming at the initial session time window, correcting the start and stop positions of each initial session time window, and obtaining the session time window.
4. 4. The method for video compression and intelligent reconstruction with very low bandwidth as in claim 3, wherein said generating a semantic recoverable hierarchical graph comprises the steps of, Based on a session time window, combining video frames in the window into a plurality of space-time areas according to fixed space division and fixed time span, and generating space-time area identifiers for the space-time areas; according to the space-time region identification, statistics of adjacent frame prediction residual variation, motion variation degree, texture randomness degree, edge structure continuity, contour stability and brightness consistency are carried out on each space-time region, so as to form a generatability statistics record and a constraint statistics record; Based on the constraint statistical record, collecting the existence of the target category, the appearance condition of the characters and the coverage proportion of the task related area for each space-time area to form a semantic key strength record; based on the generatability statistics record, the constraint statistics record and the semantic criticality intensity record, the space-time areas are subjected to joint judgment and mapped into different semantic recoverable grades, and a semantic recoverable grade diagram is generated according to the space position and the time position.
5. 5. The method for video compression and intelligent reconstruction with very low bandwidth as in claim 4, wherein the generating of the guaranteed-bottom content data and the enhanced description data comprises the following steps, According to the semantic recoverable level diagram, performing level marking according to each space-time region, and marking into a space-time region with a low recoverable level and a space-time region with a high recoverable level; for the space-time region with low recoverable grade, extracting basic pixel information, contour information and motion trend information from an original video frame sequence, and collecting and forming bottom-protecting content data; for the space-time region with high recoverable grade, extracting structural guide information, local residual information and semantic guide information from the original video frame sequence, and converging to form enhancement description data.
6. 6. The method for video compression and intelligent reconstruction with very low bandwidth as in claim 5, wherein the forming of the guaranteed bit stream and the enhanced bit stream comprises the following steps, Respectively carrying out data arrangement and field organization on the bottom-protecting content data and the enhanced description data, and carrying out structural arrangement on various information according to time sequence and space position to form normalized input data; Based on normalized input data, under the constraint of a target extremely low bandwidth, performing compression coding processing with protected priority on the bottom-protected content data to form a bottom-protected code stream; and performing self-adaptive compression coding processing on the enhancement description data according to the residual available bandwidth after the occupation of the base protection code stream is deducted, so as to form an enhancement code stream.
7. 7. The method for very low bandwidth video compression and intelligent reconstruction as set forth in claim 6, wherein said semantic guidance information is auxiliary description information extracted from an original video frame sequence and transmitted with enhancement description data.
8. 8. The method for very low bandwidth oriented video compression and intelligent reconstruction as set forth in claim 7, wherein said generating a target quality video sequence comprises the steps of, The decoding end receives the bottom protection code stream and the enhancement code stream, and preferentially executes decoding processing on the bottom protection code stream to obtain a basic video sequence; Based on the basic video sequence, performing constrained intelligent reconstruction processing on a corresponding space-time region in the basic video sequence according to structural guide information, local residual information and semantic guide information carried in the enhanced code stream; And fusing and correcting consistency of each space-time region after intelligent reconstruction processing according to the spatial position and the time sequence to form a target quality video sequence.
9. 9. The method for video compression and intelligent reconstruction with very low bandwidth as in claim 8, wherein said adaptively updating semantic restorable ranking rules and code rate allocation policies in a session time window is performed as follows, Counting the identification consistency, time sequence stability and reconstruction continuity of the corresponding region in the target quality video sequence in a session time window to form input information; according to the input information, executing task completion evaluation processing on the target quality video sequence of the current session time window to obtain task completion evaluation information; correcting the generatability statistical record, the constraint statistical record and the semantic criticality intensity record according to the task completion degree evaluation information to set a semantic recoverable grade classification rule; According to the task completion evaluation information, performing self-adaptive adjustment on semantic recoverable grade classification rules, and re-executing semantic recoverable grade judgment on video content to obtain corresponding semantic recoverable grade distribution; And carrying out self-adaptive updating on the code rate allocation strategy of the bottom protection code stream and the enhancement code stream based on the semantic recoverable grade distribution.
10. 10. The video compression and intelligent reconstruction system oriented to the extremely low bandwidth is based on the video compression and intelligent reconstruction method oriented to the extremely low bandwidth, which is characterized by comprising a session construction module, a hierarchical organization module, a code rate allocation module, an intelligent reconstruction module and a feedback update module; the session construction module is used for collecting an original video frame sequence and constructing a session time window, performing semantic analysis of video coding on video content in the session time window and generating a semantic recoverable grade diagram; The hierarchical organization module is used for dividing the video content into video areas with different recoverable grades according to the semantic recoverable grade diagram, and performing hierarchical organization on the video content to generate bottom-protected content data and enhancement description data; the code rate distribution module is used for executing content self-adaptive code rate distribution on the bottom-protecting content data and the enhancement description data under the constraint of a target extremely low bandwidth and respectively carrying out compression coding to form a bottom-protecting code stream and an enhancement code stream; The intelligent reconstruction module is used for receiving the bottom protection code stream and the enhancement code stream at the decoding end, decoding the bottom protection code stream, and simultaneously executing intelligent reconstruction processing according to the corresponding recoverable grade constraint by combining semantic guide information carried in the enhancement code stream to generate a target quality video sequence; and the feedback updating module is used for performing task completion degree evaluation based on the target quality video sequence and carrying out self-adaptive updating on the semantic recoverable grade dividing rule and the code rate allocation strategy in the session time window.

Description

Video compression and intelligent reconstruction method and system for extremely low bandwidth Technical Field The invention relates to the technical field of video processing, in particular to a video compression and intelligent reconstruction method and system for extremely low bandwidth. Background With the development of video perception and intelligent analysis technology, video data is increasingly widely applied to remote monitoring, intelligent perception and low-bandwidth communication scenes, the existing video compression technology is mainly developed around mechanisms such as pixel redundancy elimination, motion prediction, transform coding and the like, video frames are uniformly coded and decoded to complete transmission and restoration of video content, video is generally treated as a continuous image sequence to be processed, compression efficiency and reconstruction quality of a signal layer are focused, and basic video transmission and display requirements can be met under the condition of conventional bandwidth. Under the condition of extremely low bandwidth, different areas in video content have obvious differences in terms of semantic importance, structural stability and reconstructable difficulty, the traditional unified coding mode is difficult to effectively describe and utilize the differences, key area information is easy to lose or reconstruction quality is easy to decline, further the subsequent recognition, analysis and task execution effects are influenced, and in a scene needing to consider both video understandability and task availability, the prior art lacks a processing mechanism capable of comprehensively describing the video content generatable characteristic, structural constraint characteristic and semantic importance degree, and is difficult to realize a stable and controllable video reconstruction effect under the condition of extremely low bandwidth. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a video compression and intelligent reconstruction method oriented to extremely low bandwidth, which solves the problem that video key semantic information is difficult to stably reserve and effectively reconstruct under the extremely low bandwidth condition. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides a video compression and intelligent reconstruction method for extremely low bandwidth, which comprises the steps of collecting an original video frame sequence, constructing a session time window, executing semantic analysis processing of video coding on video content in the session time window to generate a semantic recoverable grade diagram, dividing the video content into video areas of different recoverable grades according to the semantic recoverable grade diagram, carrying out hierarchical organization on the video content to generate bottom-protected content data and enhanced description data, executing content self-adaptive code rate allocation on the bottom-protected content data and the enhanced description data under the constraint of an extremely low bandwidth of a target, respectively carrying out compression coding on the bottom-protected content data and the enhanced description data to form a bottom-protected code stream and an enhanced code stream, receiving the bottom-protected code stream and the enhanced code stream at a decoding end, simultaneously carrying out intelligent reconstruction processing according to the constraint of the corresponding recoverable grade by combining semantic guiding information carried in the enhanced code stream to generate a target quality video sequence, executing task completion degree assessment on the basis of the target quality video sequence, and carrying out self-adaptive updating on a semantic recoverable grade dividing rule and an allocation strategy in the session time window. As an optimal scheme of the video compression and intelligent reconstruction method facing the extremely low bandwidth, the original video frame sequence comprises multi-frame video image data and time identification information. As a preferable scheme of the video compression and intelligent reconstruction method facing to the extremely low bandwidth, the invention constructs a session time window, and comprises the following specific steps, Sequentially traversing the original video frame sequence according to the time identification information, and calculating brightness change, chromaticity change and spatial structure change between adjacent video frames frame by frame to form a content change measure; acquiring a change judgment threshold value based on the statistical distribution characteristic of the content change measurement, judging the position of the content change measurement exceeding the change judgment threshold value as a session bound