Binocular 4K picture quick browsing method

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a binocular 4K picture quick browsing method, which comprises the following steps of: 1, reading a touch coordinate and touch time from system touch callback, and entering a quick browsing state if two continuous dragging intensities reach a high-grade threshold value; otherwise, maintaining a common browsing state; 2, according to a browsing state judgment result, continuously advancing a duty cycle, and executing foresight replay and level rollback operation when triggering after time delay or packet loss is detected; 3, when it is detected that the dragging strength is reduced to a low-level threshold value and lasts for a preset time length, or when a touch ending event is detected, the current preview duty sequence is frozen, and a stable stopping replacement and progressive recovery process is entered. According to the method, the problems of display lag, binocular asynchronization, unrecoverable packet loss and large stop switching delay in the prior art are solved, and the interaction performance and the practical value of the ultra-high-definition image on middle-end and low-end equipment are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data processing, and particularly relates to a binocular 4K picture fast browsing method. BACKGROUND

[0002] With the continuous development of display devices and image acquisition technology, ultra-high definition images (Ultra High Definition, referred to as UHD) gradually become an important application form in the fields of industrial detection, virtual reality, medical imaging, security monitoring and consumer electronics. Especially in the binocular imaging and panoramic stitching scene, the resolution of a single image often reaches 4K or even 8K, thereby providing more detailed information for the human eye or computer vision system. However, in actual application, the browsing and interaction of ultra-high definition images are still limited by the hardware performance of terminal devices, especially on Android devices with insufficient computing power and coding and decoding efficiency, it is still difficult to quickly browse ultra-high definition images.

[0003] On a medium and low-end Android device, when a user quickly drags an ultra-high definition image, the system often cannot complete the high-resolution decoding and rendering of the target viewport area in real time after each touch event occurs. This causes a significant delay between the user's dragging action and the image display, commonly known as "image not following hand". Especially in binocular mode, the synchronous rendering delay of the two channels will further amplify this discomfort, seriously affecting the interactive experience. When a user browses an ultra-high definition image, there are usually two types of needs: one is to quickly locate a region in the image, which usually requires low-resolution or sampling to achieve fast dragging; the other is to accurately analyze the defects of the local region after positioning, which requires a complete 4K original image. However, the existing technology either can only guarantee fast preview and cannot consider detailed analysis, or sacrifices browsing smoothness when pursuing details, lacking a dynamic switching mechanism that considers both types of needs. SUMMARY

[0004] In view of this, the main purpose of the present application is to provide a binocular 4K picture fast browsing method, which realizes dynamic switching between fast browsing and ordinary browsing through browsing state determination, reduces rendering pressure and maintains binocular synchronization during fast dragging by using duty cycle sequence, hierarchical view index and lockstep mechanism, and solves the tile missing and delay problem by combining cross-eye image compensation, forward-looking compensation and packet loss self-healing technology; after the user stops, the gradual original level backfilling is used to restore high-definition display in a ring, ensuring the integrity of the details. Thus, the present application takes into account the needs of fast and smooth browsing experience and accurate defect analysis, overcomes the problems of display lag, binocular asynchronization, unrecoverable packet loss and large delay in switching after stopping in the prior art, and significantly improves the interactive performance and practical value of ultra-high definition images on medium and low-end devices.

[0005] The technical solutions adopted by the application are as follows: A binocular 4K picture fast browsing method, the method comprising: Step 1: reading touch coordinates and touch time from a system touch callback, buffering at least three continuous touch callbacks in chronological order, calculating the displacement and interval corresponding to the drag intensity of the adjacent two touch callbacks, if the drag intensity of the continuous two times reaches a high level threshold, entering a fast browsing state, otherwise maintaining a normal browsing state; Step 2: according to the browsing state determination result, constructing a hierarchical view index and generating a tile fingerprint chain covering each level tile; generating a preview duty sequence composed of a reserved mark, a skip mark and a forward-looking mark according to the browsing state, the preview duty sequence containing a binocular lockstep mark for synchronizing the left and right eye channels; selecting the view level and the viewport related tile set synchronously according to the duty position and the binocular lockstep mark, rearranging and losslessly packaging the selected tiles and sequentially transmitting them; the decoding end unpacks and renders the tiles from the left and right eye channels in pairs within the lockstep display time window, and generates a temporary preview tile by compensating for the missing tile with a cross-eye image; the duty cycle continues to advance, and the forward-looking supplement and level rollback operations are triggered when a time delay or a packet loss is detected.

[0006] Step 3: when the drag intensity is detected to decrease to a low level threshold and last for a preset length of time, or when a touch end event is detected, freeze the current preview duty sequence, enter the steady replacement and gradual recovery process.

[0007] Further, in step 3, the steady replacement and gradual recovery process is: filling the original level tiles from the center outward according to the priority tile table, replacing the temporary preview tiles marked as cross-eye sources with original source tiles one by one, and keeping the binocular lockstep mark and the lockstep display time window unchanged during the entire replacement period, until the current viewport area reaches the complete display of the original level.

[0008] Further, in step 1, the first displacement distance and the first time interval are calculated based on the first touch record and the second touch record; the second displacement distance and the second time interval are calculated based on the second touch record and the third touch record; wherein the displacement distance is calculated according to the pixel distance between the two touch coordinates in the screen pixel coordinate system; the time interval is calculated according to the millisecond difference of the two touch times; the first drag intensity is obtained by ratio calculation of the first displacement distance and the first time interval; the second drag intensity is obtained by ratio calculation of the second displacement distance and the second time interval; when any time interval is less than a preset valid time interval threshold, discard this determination and wait for the next system touch callback to complete the valid time interval.

[0009] Further, four levels of hierarchical view index are established in memory for each picture of left and right eyes, including original level, secondary level, tertiary level and quaternary level, wherein each level is composed of a grid of tiles with fixed side length, and each tile has a unique number and a fixed length fingerprint; the fixed length fingerprint is generated by the pixel content of the tile through an irreversible digest function.

[0010] Further, the fixed length fingerprints of the tiles are connected in the order of tile numbers to form a tile fingerprint chain.

[0011] Further, the process of generating a preview duty sequence according to the browsing state includes: selecting a preview duty sequence template according to the browsing state determination result, wherein the duty sequence template is arranged in a cyclic order by three types of markers, including a reserved marker, a skip marker and a forward-looking marker; when in a fast browsing state, a duty sequence template with a skip marker proportion exceeding a set skip threshold is selected; when in a normal browsing state, a duty sequence template with a reserved marker proportion exceeding a set reserved threshold is selected; each duty sequence template includes a binocular lock step marker, which is used to ensure that the left and right eye channels perform consistent duty relationships in the same cycle.

[0012] Further, the tile set obtained from the lock step selection is rearranged and losslessly packaged according to the tile fingerprint chain as a sequential reference; each packaged segment includes a hierarchical marker header, a tile number, a display order number and a tile data area, the hierarchical marker header of the packaged segment indicates the view level and whether the packaged segment comes from the forward-looking marker, and the display order number is used to restore the display order according to the duty position at the decoding end.

[0013] Further, the decoding end pairs the packaged segments of the left and right eye channels according to the display order number, and renders the segments with the same duty position in the lock step display time window; if a segment is missing in a channel in the current time window, a cross-eye image compensation process is entered; when a target tile is missing in the current duty position in a channel, a tile with the same duty position, the same view level and the same tile number is obtained from the other channel to generate a temporary preview tile; the generation method is: copying the obtained tile, performing geometric translation according to a preset fixed horizontal offset, performing boundary clipping and pixel-level splicing with adjacent tiles; the temporary preview tile is marked as a cross-eye source and can be overwritten by a subsequent original source tile; in the duty position corresponding to the forward-looking marker, the forward-looking tile from the quaternary view is preferentially rendered to reduce visible delay; if the decoding end detects that the same tile does not arrive for two consecutive time windows, the view level of the tile is lowered by one level and is requested again, until it is filled in the subsequent time window or enters the step three stable replacement process.

[0014] Further, in step 2, after completing an empty cycle, it is decided according to the current browsing state whether to continue using the same preview empty sequence or switch to a template with higher reserved mark ratio; when the rendering time exceeds the lockstep display time window, the view level of the next cycle is immediately uniformly downgraded by one level.

[0015] By adopting the above technical solutions, the present application has the following beneficial effects: the smoothness of image browsing and the accuracy of detail analysis can be maintained on terminal equipment with limited hardware performance. First, the browsing state determination mechanism can automatically identify the fast browsing state and the ordinary browsing state according to the user's dragging intensity, and dynamically switch the preview empty sequence in different states, so that the image display can ensure smooth experience during interaction and gradually restore high-resolution images after stopping, solving the problems of image display delay and dragging out of sync in the traditional method. Second, the binocular lockstep interlaced emptying mechanism ensures that the left and right channels execute consistent preview strategies in the same cycle, thereby avoiding stereoscopic vision distortion caused by different steps of binocular display, and generating temporary preview tiles in real time when a tile is missing in a certain channel through cross-channel image compensation technology, effectively improving the consistency and stability of binocular display. Third, by introducing the forward-looking patching and packet loss self-healing mechanism, the tiles in the future viewport direction can be rendered in advance when the user moves quickly, significantly reducing the user's perceived delay, and automatically downgrading the request for lower-level tiles when tiles are continuously missing, avoiding the problem of blank areas or long waiting. Finally, when the user stops dragging or slows down, the current preview sequence can be frozen, and the stopping replacement and gradual recovery process can be started, the original level tiles are gradually filled according to the priority tile table, and the center to the edge is covered in a ring, which ensures that the user can immediately see the clear image of the core area, and gradually restores the full-resolution display without increasing the delay. This gradual replacement and recovery mechanism enables the user to smoothly locate the image defect position during fast browsing and seamlessly enter the fine analysis state after stopping, meeting the dual needs of speed and quality. Through the above technical means, the present application effectively overcomes the problems of browsing lag, binocular asynchronization, unresolvable packet loss and large delay in the prior art, and realizes efficient, smooth and sustainable binocular 4K picture browsing on devices with limited performance. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The method flowchart of the binocular 4K picture fast browsing method provided by the embodiment of the present application is shown in the figure. Figure 2 The tile transmission and packet loss rate analysis curve diagram provided by the embodiment of the present application is shown in the figure. Figure 3 The empty sequence switching and transmission efficiency analysis diagram provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0017] All features disclosed in this specification, or all steps in the disclosed methods or processes, except mutually exclusive features and / or steps, can be combined in any manner.

[0018] Any feature disclosed in this specification (including any appended claims and abstract), unless otherwise stated, may be replaced by other equivalent or similar features. In other words, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0019] refer to Figure 1 : A binocular 4K image fast browsing method, the method comprising: Step 1: Read the touch coordinates and touch time from the system touch callback, cache at least three consecutive touch callbacks in chronological order, calculate the drag strength corresponding to the displacement and interval between two adjacent touch callbacks, and if the drag strength reaches the high level threshold for two consecutive times, enter the fast browsing state; otherwise, maintain the normal browsing state.

[0020] During the specific implementation process, a touch record buffer queue with a capacity of 3 is established, storing the first touch record, the second touch record, and the third touch record in chronological order. Each touch record contains only the touch coordinates in the screen pixel coordinate system and the touch time in milliseconds. When the system touch callback generates a new mobile touch event, the touch coordinates and touch time corresponding to the event are written to the third touch record, and the original third touch record is moved down to the second touch record, and the original second touch record is moved down to the first touch record. If the buffer queue has less than 3 records, no judgment is made and the system continues to receive touch records until the buffer queue contains 3 valid touch records.

[0021] The first displacement distance and the first time interval are calculated based on the first touch record and the second touch record; the second displacement distance and the second time interval are calculated based on the second touch record and the third touch record. The displacement distance is calculated as the pixel distance between the two touch coordinates in the screen pixel coordinate system; the time interval is calculated as the millisecond difference between the two touch times. The first displacement distance is calculated by the ratio of the first time interval to obtain the first drag strength; the second displacement distance is calculated by the ratio of the second time interval to obtain the second drag strength. To avoid abnormal jitter affecting the judgment, when any time interval is less than the time resolution capability of the system touch callback, the current judgment is discarded and the next system touch callback is waited for to complete the effective time interval.

[0022] A high-level threshold is set in advance in the configuration storage, which is used to distinguish between normal browsing state and fast browsing state. After the calculation of the drag intensity is completed twice each time, the first drag intensity and the high-level threshold are compared in turn, and the second drag intensity and the high-level threshold are compared: when the first drag intensity is greater than or equal to the high-level threshold and the second drag intensity is greater than or equal to the high-level threshold, the browsing state flag is set to the fast browsing state; in any other case, the browsing state flag is set to the normal browsing state. To ensure the certainty of the determination, the browsing state flag is output only once in a system touch callback processing flow, and the browsing state flag is immediately provided to the subsequent steps.

[0023] If the touch coordinates are missing, the touch time is missing, or one of them is invalid, the corresponding touch record is cleared and the browsing state flag is kept as the normal browsing state; if the system touch callback is temporarily interrupted, causing the buffer queue time span to abnormally increase, the browsing state flag is not updated this time, and the previous browsing state flag is used; when a new moving touch event is first received after detecting the touch end event, the touch record buffer queue is reset, and the above determination process is resumed only after the new buffer queue accumulates 3 valid touch records.

[0024] Step 2: According to the browsing state determination result, a hierarchical view index is constructed and a tile fingerprint chain covering tiles at all levels is generated; a preview blanking sequence composed of a reserved mark, a skip mark and a forward-looking mark is generated according to the browsing state, the preview blanking sequence includes a binocular lock step mark for synchronizing the left and right eye channels; according to the blanking position and the binocular lock step mark, the view level and the viewport related tile set are selected synchronously, the selected tiles are rearranged and losslessly packaged and are sequentially transmitted; the decoding end pairs and unpacks the tiles from the left and right eye channels within the lock step display time window, generates a temporary preview tile to compensate for missing tiles with cross-eye mirroring; the blanking cycle continues to advance, and the forward-looking supplement and level rollback operations are triggered when a time delay or a packet loss is detected.

[0025] In the implementation process, the upper left corner of the screen is taken as the coordinate origin, the horizontal direction to the right is taken as the horizontal positive direction, the vertical direction downward is taken as the vertical positive direction, and all coordinates are represented in pixels. The pixel arrangement of each image is unified as the order of the red channel, the green channel, the blue channel and the transparency channel, the single-channel bit depth is 8 bits, and the pixels are stored from top to bottom and from left to right. Four levels of original, secondary, tertiary and quaternary are established for each image. The original level directly references the input pixels. The secondary level is generated by integer scaling from the original level, and the scaling rule is to reduce the horizontal and vertical pixel numbers to half of the original level. The pixel sampling adopts a fixed strategy of block mean approximation or adjacent pixel replacement and maintains a unique option in the configuration storage. The tertiary level is generated from the secondary level with the same rule. The quaternary level is generated from the tertiary level with the same rule. After each level is generated, it is immediately frozen and no longer modified in subsequent steps.

[0026] The same fixed-length tile length is set for the four levels, the fixed-length tile length is stored in the configuration storage and represented in pixels, for example, 256. Each level is evenly divided into a grid along the horizontal and vertical directions according to the fixed-length tile length. When the rightmost or lowermost side is less than the fixed-length tile length, the pixel buffer is padded with zero values to make each level consist of a complete grid of fixed-length tiles. Each fixed-length tile of each level is sequentially numbered from top to bottom and from left to right. The unique number is obtained by sequentially splicing four parts, namely the image identifier, the level number, the row number and the column number; the image identifier is used to distinguish between left and right channels, the level number is assigned to the original, secondary, tertiary and quaternary levels as 1, 2, 3 and 4, and the row number and the column number are sequentially incremented from 1. The unique number is not repeated in the same browsing session of the same terminal. A fixed-length fingerprint is generated for each fixed-length tile, and the fixed-length fingerprint length is 16 bytes. The generation steps are as follows: without changing the pixel value, read the pixel data of the fixed-length tile, connect the red channel byte, the green channel byte, the blue channel byte and the transparency channel byte in sequence to form a continuous byte stream in row priority order; the padding area is reserved with zero value bytes and is involved in the connection; the continuous byte stream is input into an irreversible digest function, and a 16-byte result is output; the 16-byte result is taken as the fixed-length fingerprint of the fixed-length tile. The irreversible digest function is uniquely specified in the configuration storage and remains unchanged throughout the system life cycle. The fixed-length fingerprint is only used for fast consistency check and cross-frame reuse determination, and is not used for image display and color operation.

[0027] A level description record is established for each level, which contains a width measured in pixels, a height measured in pixels, a number of grid rows, a number of grid columns, a fixed side length tile side length, a number of fixed side length tiles, and a start position of a fixed side length tile description record array. A fixed side length tile description record is established for each fixed side length tile, which contains a unique number, a level sequence number, a row sequence number, a column sequence number, a top-left corner coordinate measured in pixels, a bottom-right corner coordinate measured in pixels, an offset position of pixel data in a pixel buffer, a pixel data length, and a fixed length fingerprint. The level description records of the four levels are sequentially written into a hierarchical view index table header, and all the fixed side length tile description records are sequentially written into a hierarchical view index table body, to form a complete hierarchical view index. After writing is completed, each fixed side length tile description record is sequentially traversed, and the fixed length fingerprint is used to perform consistency checking on multiple generation results of the same unique number in the current session; when consistent, it is marked as reusable, and when inconsistent, it is marked as non-reusable and a timestamp is recorded. The hierarchical view index is stored in memory and can be optionally written to persistent storage, and when written, the header is kept in front, the body is kept behind, and is saved in a page-aligned manner, and when read, it is restored in the same order.

[0028] Further, the preview duty cycle sequence is selected and generated only according to the browsing state determination result, and is bound with the binocular lockstep flag during generation to ensure consistent duty cycle relationship between the two channels, including: two types of template sets are pre-stored in the configuration storage, which are fast browsing state template set and ordinary browsing state template set. Each template consists of template identifier, cycle length, duty cycle flag array, template version number and creation time, wherein the duty cycle flag array is arranged in a fixed order by three types of flags, i.e. reserved flag, skip flag and forward-looking flag. The proportion of skip flag in each template of fast browsing state template set is higher than that of reserved flag, and the proportion of reserved flag in each template of ordinary browsing state template set is higher than that of skip flag. When the browsing state flag is fast browsing state, a template with cycle length of 8 and containing at least 4 skip flags is selected from the fast browsing state template set; when the browsing state flag is ordinary browsing state, a template with cycle length of 12 and containing at least 8 reserved flags is selected from the ordinary browsing state template set; if there is a template being used in the previous cycle and the browsing state has not changed, the template is continued to be used until the end of the cycle.

[0029] A binocular lockstep marker is generated for the selected template, which includes a template identification, a loop counter initial value, a blank position sequence number initial value, and a left and right channel synchronization sequence number. The binocular lockstep marker is written into the synchronization buffer of the left and right channels at the same time, and the blank position sequence number is set to 1 in both channels. At the beginning of each rendering cycle, the blank position sequence number in the binocular lockstep marker is read, and the marker type of the current blank position is determined according to the blank marker array; after the current blank position processing is completed, the blank position sequence number is incremented by 1; when the blank position sequence number exceeds the loop length, the blank position sequence number is reset to 1 and the loop counter is incremented by 1; the order and content of the blank marker array must not be modified during the entire loop.

[0030] Lockstep selection and view level determination, based on the binocular lockstep marker, select the view level and tile set for preview from the hierarchical view index for each blank position, and perform fixed rules corresponding to the reserved marker, the skip marker, and the forward-looking marker, respectively, as follows: obtain the viewport rectangle provided by the current rendering system, and locate the tile set intersecting the viewport in the hierarchical view index as the viewport tile set with a fixed side length tile side length as a step; add the outer ring tiles sharing edges or corners with the viewport tile set to form a ring neighborhood tile set; merge the two into the viewport and ring neighborhood tile set. Maintain a last reserved tile set cache for each channel for direct reuse at the blank position corresponding to the skip marker; when executing the blank position corresponding to the reserved marker, overwrite the last reserved tile set cache with the new viewport and ring neighborhood tile set. When in fast browsing state, the target view level is set to tertiary; when in normal browsing state, the target view level is set to secondary; when the target view level is not fully loaded before the end of a loop, temporarily reduce the target view level to tertiary at the blank position corresponding to the first reserved marker of the next loop until loading returns to normal. Read the target view level and the viewport and ring neighborhood tile set, and extract the fixed side length tiles matching the tile set in the corresponding level of the hierarchical view index in the order of unique number as the preview tile set of the current blank position; write the preview tile set into the last reserved tile set cache for subsequent skip marker reuse. Do not add new tiles from the hierarchical view index, but directly reference the last reserved tile set cache as the preview tile set of the current blank position; if the last reserved tile set cache is empty, immediately perform a reserved marker processing and write the result into the cache before continuing to advance the blank position sequence number.

[0031] According to the latest three touch records, the latest moving direction is calculated, specifically: comparing the difference in horizontal and vertical coordinates of the third touch record and the second touch record in sign and absolute value, the moving direction is discretized into one of right, left, down, up, right down, right up, left down and left up; taking the outer edge of the viewport as the starting point, sequentially selecting continuous tiles on the grid of the target view level along the moving direction, generating a front view tile queue with a length of 12; in the empty position corresponding to each front view mark, 4 tiles are taken from the head of the front view tile queue as the preview tile set of the empty position and removed from the queue; when the moving direction is detected to be inconsistent with the last empty position, the front view tile queue is emptied and regenerated according to the new moving direction. The left and right channels read the same binocular lock step mark and empty mark type at the same empty position, and complete the reserved mark processing, skip mark processing or front view mark processing according to the above rules to obtain consistent preview rhythm of the two channels; when any channel fails to complete tile selection on time at the current empty position, the two channels simultaneously use the last reserved tile set to maintain consistency in display.

[0032] The tile fingerprint chain construction and lossless packaging includes the following specific implementation process. In each occupancy position, the sending end performs the following steps according to the binocular lockstep mark: reading the tile set selected by the lockstep in the current occupancy position, each fixed-length tile in the tile set having a unique number and a fixed-length fingerprint. The tile set is sorted in ascending order of the unique number as the unique sequence reference for subsequent concatenation and packaging. The fixed-length fingerprints are read in order according to the sorted unique number, and the continuous byte sequence is sequentially connected to obtain the tile fingerprint chain. To facilitate consistent and fast verification at the unpacking end, the first fixed-length fingerprint is extracted from the tile fingerprint chain as the fingerprint chain start fingerprint, the last fixed-length fingerprint is extracted as the fingerprint chain end fingerprint, and an irreversible digest function is used to calculate the packaging check code for the entire tile fingerprint chain. The fingerprint chain start fingerprint, the fingerprint chain end fingerprint and the packaging check code are only used for consistency check and do not participate in image display. Each packaging segment is sequentially composed of a hierarchical mark header, a tile number, a display sequence number, a fingerprint chain start fingerprint, a fingerprint chain end fingerprint, a packaging check code and a tile data area. The hierarchical mark header includes a view level field and a previous view source field, the view level field takes one of the original level, secondary level, tertiary level or fourth level, and the previous view source field takes one of the previous view mark or non-previous view mark. The tile number is the count of fixed-length tiles contained in the current packaging segment. The display sequence number is a decimal number obtained by sequentially splicing the current loop counter and the current occupancy position sequence number, which is used to restore the display order according to the occupancy position at the receiving end. The tile data area contains a plurality of tile records, each tile record sequentially contains a unique number, a fixed-length fingerprint, a pixel data length and pixel data, and the pixel data is written row by row according to the order of red channel, green channel, blue channel and transparency channel. All fields are written in byte alignment, and the segment tail is aligned to 4 bytes with zero padding.

[0033] According to the order of the tile fingerprint chain, the tile records in the tile data area are sequentially written, so that the tile order of the tile data area is consistent with the tile fingerprint chain order. The pixel data is not lossy processed, and is compressed in a lossless manner or uncompressed. The specific manner is uniquely specified in the configuration storage. After completing the writing of the tile data area, the hierarchical mark header, the tile number, the display sequence number, the fingerprint chain start fingerprint, the fingerprint chain end fingerprint and the packaging check code are written to generate a complete packaging segment. The packaging segment is pushed into the sending buffer queue for transmission or local rendering pipeline consumption. When the unique number is missing, the fixed-length fingerprint is missing, or the pixel data length and the pixel data are inconsistent during the packaging process, the tile record is discarded and is not counted in the tile number field, and the record is ignored during the calculation of the packaging check code; if the tile data area is empty, an empty packaging segment containing only the hierarchical mark header, the tile number being 0 and the display sequence number is output in the current occupancy position to maintain the continuity of the display sequence.

[0034] The binocular pair end unpacking and lockstep display includes the following specific implementation process, which is executed at the receiving end for two channels respectively and synchronized through the binocular pair time queue: the receiving end reads the encapsulated fragment from the transmission layer or the local rendering pipeline, parses the hierarchical tag header, tile quantity, display sequence number, fingerprint chain start fingerprint, fingerprint chain end fingerprint, and encapsulation check code, and checks that the encapsulation check code is consistent with the fragment content; if not, mark the encapsulated fragment as an invalid fragment and discard it. For valid fragments, establish a fragment index entry, use the display sequence number as the key, and write the fragment pointer into the receiving buffer of the corresponding channel. Traverse the tile data area in sequence, read the unique number, fixed length fingerprint, pixel data length, and pixel data for each tile record, calculate the fixed length fingerprint of the pixel data using the same irreversible digest function, and compare it with the fixed length fingerprint in the record. If they are consistent, write the tile record to the tile cache; if they are inconsistent, discard the tile record and mark it as missing in the tile cache. After the traversal is completed, the local fingerprint chain is formed by sequentially connecting the fixed length fingerprints in the tile cache, and the start and end fixed length fingerprints of the local fingerprint chain are compared with the fingerprint chain start fingerprint and the fingerprint chain end fingerprint in the fragment. If they are consistent, mark it as a usable fragment; if they are inconsistent, mark it as an unusable fragment and discard it.

[0035] A reach table is maintained for the left and right two channels respectively, and the usable fragment pointer is stored as the key of the display sequence number. A binocular pair time queue is established, and the paired fragment state is stored as the key of the display sequence number. Whenever a usable fragment reaches any channel, the other channel fragment with the same display sequence number is found in the binocular pair time queue: if it exists, mark the display sequence number as paired complete; if it does not exist, record it as a half-paired state and wait for the other channel to arrive. Read the lockstep display time window length in the configuration storage, and use the display sequence number increment as the time advancing sequence. When a certain display sequence number enters the current lockstep display time window, the following rules are executed: if paired complete, take out two fragments from the binocular pair time queue, drive rendering according to the view level and previous view source field in the hierarchical tag header, write the tile records in the fragment to the current view buffer according to the unique number sequence, and perform display; if only one channel arrives, immediately enter the cross-eye image compensation process to generate temporary preview tiles for the missing channel, and complete rendering in the current time window; if neither channel arrives, reuse the last reserved tile set cache to complete rendering and record a frame loss event.

[0036] When the receiving end receives a segment with display order number less than the lower bound of the current lockstep display time window, it is marked as expired and discarded; when a segment with the same display order number as the rendered number is received, it is marked as repeated and discarded; when a segment with a display order number greater than the upper bound of the current lockstep display time window is received, it is retained in the receiving buffer waiting for the corresponding time window to arrive, and no early rendering is triggered. If any channel is missing available segments within the lockstep display time window corresponding to a certain display order number, the cross-eye image compensation process is called: read the available segments of the other channel at the same display order number, take out the tile records one by one and generate temporary preview tiles according to the preset fixed lateral offset, and write them into the current view buffer of the missing channel; the temporary preview tiles are marked as cross-eye source and will be overwritten by the corresponding original source tile when the missing channel's available segment is received later. After completing the paired rendering of a certain display order number, delete the corresponding entries from the binocular time queue and receiving buffer, release the tile cache and segment memory, and proceed to the next display order number to continue lockstep display time window management. If multiple display order numbers are continuously detected to trigger the frame loss event within their time windows, send a level rollback indication to the upstream to reduce the view level and shorten the arrival delay in the subsequent duty cycle.

[0037] Cross-eye image compensation, including the following specific implementation process, is executed within the lockstep display time window in the order of display order number as time advancement, and after completion, the temporary preview tile is output to the current view buffer of the missing channel: read the binocular time queue and receiving buffer, and determine whether there is a missing channel available segment or an available segment but a missing target fixed-length tile at the duty position corresponding to the current display order number; in either case, record the missing channel identifier, display order number, view level and unique number, and enter the cross-eye image compensation process. In the receiving buffer of the other channel, search for available segments by display order number as the key; in the tile data area of the available segment, find the fixed-length tile with the same view level and unique number as the missing channel; if found, read its fixed-length fingerprint and run the same irreversible digest function on its pixel data to verify the consistency of the fixed-length fingerprint; if consistent, copy the pixel data of the fixed-length tile to the working buffer and mark it as the source tile; if not found or inconsistent, terminate this compensation and use the last retained tile set cache to complete the rendering of the current duty position.

[0038] The self-configuration storage reads the fixed lateral offset and direction mapping rule, wherein the left channel is applied to right translation when missing, and the right channel is applied to left translation when missing. The pixel-level translation is performed on the source tile in the working buffer row by row, the pixels outside the target rectangle are discarded, the vacancy pixels generated after the shift are filled with zero values; the red channel, the green channel, the blue channel and the transparency channel are processed independently according to the channel and the channel order is kept unchanged. According to the upper left corner coordinate and the lower right corner coordinate of the target fixed side length tile in the missing channel current view buffer, the source tile after geometric translation is cropped to the range completely falling into the target rectangle; the pixel channel number and the channel order are not changed in the cropping process; the intermediate tile with the same size as the target fixed side length tile is obtained after the cropping is completed. The adjacent tiles sharing the upper boundary, the lower boundary, the left boundary and the right boundary with the target fixed side length tile are queried in the missing channel current view buffer; for the shared boundary with the adjacent tile, a 2-pixel wide stitching belt is established: one side from the intermediate tile and the other side from the adjacent tile; the transition processing of the channel is performed on the corresponding pixels in the stitching belt, the processing rule is that the red channel, the green channel, the blue channel and the transparency channel of the two sides of the pixels are added respectively and then divided by 2 to obtain an integer result after rounding, and the result is used to replace the pixel value of the corresponding position of the intermediate tile; for the boundary without the adjacent tile, no stitching belt is established and no processing is made.

[0039] The pixel-level stitched intermediate tile is written into the current view buffer of the missing channel, covering the target fixed-size tile position; cross-view source metadata is generated for this position in the tile cache, including display order number, view level, unique number, cross-view source channel identifier, fixed horizontal offset, and generation time; at the same time, the position is marked as a temporary preview tile for subsequent replacement. During the subsequent display order number advancement, when the missing channel receives an original source tile with the same view level and unique number as the temporary preview tile, first calculate the fixed-length fingerprint of its pixel data and compare it with the fixed-length fingerprint recorded in the tile cache; when consistent, directly overwrite the temporary preview tile with the pixel data of the original source tile and clear the cross-view source metadata; when inconsistent, keep the temporary preview tile and record an inconsistency event in the receiving buffer, waiting for the next available original source tile to arrive and repeat the consistency recovery. If steps two to six are not completed before the end of the current lockstep display time window, abandon this cross-view mirror compensation and use the last retained tile set cache for rendering, and record a compensation timeout event in the synchronization queue; when three consecutive compensation timeout events are detected, send a level rollback indication to the upstream to shorten the subsequent segment arrival delay. After completing the temporary preview tile writing, release the memory occupied by the working buffer, update the pairing status of the current display order number in the binocular synchronization queue, and continue to the next display order number to execute the lockstep display time window management; when the consistency recovery of the coverage rule is completed, clear the cross-view source metadata of the corresponding position and re-mark the position as an original source state.

[0040] The front-view compensation and packet loss self-healing includes the following specific implementation process, which is only enabled when the occupancy position of the occupancy marker is marked as a front-view marker, and is executed within the lockstep display time window: read the occupancy position number in the binocular lockstep marker, and when the current occupancy marker is determined to be a front-view marker, obtain a predetermined number of front-view tiles from the front-view tile queue in the head order, set the view level of each front-view tile to four, and mark the front-view source field in the encapsulated segment header as coming from the front-view marker to ensure that it is written into the current view buffer first after reaching the decoding end. When the display order number enters the lockstep display time window, the receiving end unpacks the paired segments, processes the tile records marked as coming from the front-view marker and with a view level of four in the header level marker first, and writes them into the corresponding position of the current view buffer in the unique number order; if there is an old tile in the same position in the last retained tile set cache, the four-level front-view tile covers the old tile.

[0041] The receiving end maintains a packet loss count entry for each unique number in the binocular synchronization queue. When a unique number does not arrive in the available segment within the current lockstep display time window of the channel and the unique number also does not arrive within the previous lockstep display time window, the packet loss count of the unique number is incremented by 1. When the packet loss count reaches 2, the view level is triggered to downgrade and re-request. For the triggered unique number, the current target view level is determined, and the target view level is downgraded according to a fixed rule: the original level is downgraded to the secondary level, the secondary level is downgraded to the tertiary level, the tertiary level is downgraded to the quaternary level, and the quaternary level is not downgraded any more. A re-request record is generated and sent to the upstream, requesting the tile record corresponding to the unique number to be retransmitted at the target view level after the downgraded level in the subsequent blank position. If the available segment of the unique number is received in any subsequent lockstep display time window, the tile record in the arrived segment is directly used to overwrite the temporary content in the same position in the current view buffer; if the position is a temporary preview tile generated by the cross-eye image compensation or a low-level forward-looking tile, the tile record in the arrived segment is used to overwrite the temporary content unconditionally.

[0042] If the above supplement operation cannot be completed before the end of the current lockstep display time window, the current view buffer content is continued to be displayed in the time window, and the time window is not extended. When the stable replacement in step three is entered in the subsequent cycle, the re-request and downgrade link of the unique number are terminated, and the original level tile is backfilled by the stable replacement process to complete the final recovery.

[0043] The blank cycle promotion and level rollback include the following specific implementation process, which is executed immediately after each blank cycle is completed, and is used to determine the preview blank sequence and view level strategy of the next blank cycle: read the blank position sequence number and cycle length in the binocular lockstep marker, when the blank position sequence number is promoted from the cycle length to 1, it is determined that the current blank cycle has been completed, and the cycle counter in the binocular lockstep marker is incremented by 1. Read the browsing state marker: when the browsing state marker is in the fast browsing state, continue to use the current preview blank sequence template; when the browsing state marker is in the normal browsing state, switch to the template with a higher reserved marker ratio. When the switching is completed, write the new template identifier into the binocular lockstep marker, and reset the blank position sequence number to 1, so that the two channels start from the same template and the same blank position in the next blank cycle. For each display sequence number in the just ended blank cycle, record the rendering time from entering the lockstep display time window to completing the current view buffer writing; if the rendering time is greater than the length of the lockstep display time window in any record, generate a level rollback indication, and write the indication into the rollback indication register shared by the two channels.

[0044] When the level rollback indication exists, the target view levels of the two channels are uniformly rolled back by one level at the first reserved marker position of the next blanking cycle: the original level is rolled back to the sub-level, the sub-level is rolled back to the third level, the third level is rolled back to the fourth level, and the fourth level is no longer rolled back; the target view level after the rollback is used for both reserved marker processing and forward marker processing; the skip marker processing still reuses the last reserved tile set cache, and does not generate new tiles separately. The uniform level rollback takes effect immediately after one application, and continues to act on the entire next blanking cycle; the target view level cannot be changed again before the end of the blanking cycle; when there is no record of rendering time greater than the lockstep display time window length in the next blanking cycle, the rollback indication is automatically cleared, and the system only advances according to the browsing state marker and the template selection rule in the subsequent cycle; if there is still a record of rendering time greater than the lockstep display time window length, the uniform level rollback continues to be performed according to the rule in the subsequent cycle until there is no longer a window. After completing template maintenance and switching and uniform level rollback, the metadata markers of the last reserved tile set cache of the left and right channels are updated at the same time, so that the view level field recorded is consistent with the current target view level; the entries marked as expired in the binocular synchronization queue are cleaned up to provide a consistent initial state for the lockstep display of the next blanking cycle.

[0045] Step 3: When it is detected that the dragging intensity decreases to the low-level threshold and lasts for a preset length of time, or a touch end event is detected, freeze the current preview blanking sequence, and enter the steady replacement and gradual recovery process.

[0046] In the specific execution process, the browsing state determination result is read, and when the trigger condition is met, the current preview blanking sequence is immediately frozen: the preview blanking sequence being used is copied as a frozen copy, the freeze flag is set to valid, and the blanking position sequence number increment is stopped; at the same time, the template identifier, the cycle counter and the blanking position sequence number in the binocular lockstep marker are retained, and the lockstep display time window length is not modified in any way; a steady start entry is written in the binocular synchronization queue, and the display sequence number after the entry is used as the time advancement reference for gradual recovery. The current viewport rectangle is located in the hierarchical view index, all fixed edge length tile sets intersecting the viewport rectangle are calculated, the level is limited to the original level, and the original level target set is obtained; unique number lists corresponding one-to-one to the original level target set are generated for the left and right channels respectively, as the only basis for subsequent backfilling.

[0047] A priority tile table is constructed starting from the fixed edge-length tile where the center of the viewport rectangle is located, and expanding outward in rings to cover all unique numbers of the original level target set: the ring number starts from 1 and increases by integers, and each ring contains outer tiles that share edges or corners with the previous ring; the sorting rule within the same ring is to record the unique numbers in clockwise direction starting from the upper midpoint; for edge tiles that are not completely covered by the viewport rectangle, their unique numbers are still included in the corresponding ring, keeping the sorting continuous. A backfill queue is established for each channel, and backfill request records are written in the order of the priority tile table, each containing at least the display order number, the view level being the original level, the unique number, and the target channel identifier; backfill requests for the same unique number from two channels are simultaneously enqueued to form a pair of backfill requests, to ensure that the backfill process follows the consistent rhythm of binocular lockstep marking.

[0048] When any channel receives an original level tile matching the backfill request record, it immediately calculates the fixed-length fingerprint of the pixel data and compares it with the fixed-length fingerprint recorded in the unique number record in the hierarchical view index; if they are consistent, the tile is marked as an available original source tile and written to the tile cache; if they are inconsistent, the tile is discarded and the original request is retained in the backfill queue, waiting for the subsequent retransmission to arrive. At the arrival of each lockstep display time window, available original source tiles for two channels are taken out from the tile cache in pairs according to the order of the priority tile table and written to the current view buffer of the two channels, and the following contents in the corresponding positions are always overwritten: forward-looking tiles from the fourth level, reserved marked tiles from the secondary or tertiary level, and temporary preview tiles from cross-eye image compensation; if only one channel arrives at an available original source tile at the current time window, and the other channel is missing, the arriving channel is written first and the status of the missing channel is maintained, and the pairing consistency is completed after the same unique number arrives at the subsequent time window.

[0049] For the temporary preview tile marked as cross-eye source in step two, detect whether the available original source tile corresponding to its unique number arrives; if so, immediately perform replacement, write the original source tile pixel data, and synchronously delete the cross-eye source metadata and related temporary mark at this location; if not, keep the status quo and do not interrupt the progressive writing process. If a unique number fails to arrive at the available original source tile within two consecutive lockstep display time windows, regenerate a backfill request record for this unique number and place it at the tail of the backfill queue without changing the view level, the binocular lockstep mark, or the lockstep display time window; this re-request can be repeatedly executed until the unique number is written to the original source tile or the freeze flag is invalidated due to a touch restart event. Continue the above progressive writing until all unique numbers in the original level target set are covered by original source tiles in the current view buffer of both channels; at this point, mark the start bar as complete, clear the freeze flag, and keep the current values of the binocular lockstep mark and the lockstep display time window unchanged; when a new drag intensity again reaches the high-level threshold is detected subsequently, end this process and hand over to the preview duty sequence to continue with lockstep selection and view level determination.

[0050] When pixel data reading fails, fixed-length fingerprint verification fails, or writing exceeds the boundary, skip the current time window writing for this unique number and keep its backfill request record in the backfill queue; when backfilling is complete or the process ends, release the tile buffer, work buffer, and backfill queue memory related to backfilling, and keep the current view buffer as the subsequent display baseline. Through the above process, without changing the binocular lockstep mark and the lockstep display time window, freeze the current preview duty sequence and backfill the original level tiles in strict order from the center outward in rings using the priority tile table, replace the temporary preview tiles of cross-eye source in real time until the current viewport area reaches complete display at the original level, thereby achieving stable replacement and progressive recovery.

[0051] A specific implementation process is given below for a pair of pictures for each eye running on the same terminal.

[0052] Let the original level pixel width be pixels, and the original level pixel height be pixels. The fixed side length tile side length is pixels. The level numbers represent the original level, secondary level, tertiary level, and quaternary level, respectively. The pixel width and pixel height of the level are and , respectively. According to the integer scaling rule, . Accordingly, . The grid row number and column number of the level are and . Substituting the calculation: . Each fixed-size tile is uniquely identified by a quadruple . Where is the channel identifier (left channel , right channel ), is the row number (from top to bottom, starting from 1), is the column number (from left to right, starting from 1). For easy indexing, a decimal unique number is constructed. Each fixed-size tile generates a fixed-length fingerprint (length 16 bytes), denoted as , which is obtained by applying an irreversible digest function to the pixel byte stream of the tile.

[0053] The lockstep display time window length is denoted as (millisecond), and in this example it is . The duty cycle length is denoted as (in this example, it is ). The cycle counter is (starting from 1), and the duty position sequence number is . The display order number is defined as , where the constant 100 is chosen as a base number greater than any possible cycle length, ensuring that the numbering within the same cycle and across cycles is monotonic and does not conflict. The fixed horizontal offset is (pixels), and in this example it is . The stitching bandwidth is (pixels), and in this example it is .

[0054] The touch record triple is (pixel, pixel, millisecond), where . In this example, we have: .

[0055] The distance of the two displacements is , The time interval between the two times is .

[0056] The intensity of the two drags is .

[0057] The high-level threshold is denoted as , and in this example it is configured as . Because and , the fast browsing state is entered.

[0058] According to the previous results, the three levels are , and the four levels are . The unique tile number is The unique number of, for example, the left channel, level 3, row 2, column 3 is calculated as .

[0059] The fast browsing state selects a template with length 8, in order: p = 1 reserved marker, p = 2 skip marker, p = 3 skip marker, p = 4 look-ahead marker, p = 5 reserved marker, p = 6 skip marker, p = 7 skip marker, p = 8 look-ahead marker. Initialize the loop counter , the position sequence number of the blanking position . The first display order number of the blanking loop is .

[0060] The lockstep selection (the first blanking position, reserved marker, target view level is level 3) has the upper left corner of the viewport rectangle in the level 3 coordinate system as , and the width and height are , respectively. The column index range covered by the viewport is .

[0061] The row index range covered by the viewport is .

[0062] The viewport tile set contains 4 tiles in total. The ring neighborhood is extended to the direct product of rows and columns in this example, which is exactly the same as the entire grid of level 3, with 12 tiles in total. Therefore, 12 level 3 tiles need to be sent for each channel when p = 1. Take three example unique numbers (left channel): .

[0063] Generate fixed-length fingerprints for the 12 tiles and form a tile fingerprint chain in ascending order of unique numbers , where the symbol " " " represents byte-level connection. Similarly, form for the right channel.

[0064] Perform rearrangement lossless packaging (p = 1): the view level field of the level marker header is recorded as , and the look-ahead source field is 0 (not from the look-ahead marker). The number of tiles is recorded as . The display order number is . Write the records in the tile data area in the order of the tile fingerprint chain. The packaging check code is recorded as , which is obtained by applying an irreversible digest function to the entire fingerprint chain and taking the first 4 bytes. For example, the header tuple of the left channel p = 1 packaging segment can be written as .

[0065] Perform receiver unpacking and lockstep display (p = 1): when the left and right channel segments of p = 1 are in the time window On arrival of the innermost tile, write the 12 tertiary tiles in ascending order of unique number to the current view buffer and render. If the right channel is missing a tile with unique number , enter crosshairs image compensation immediately.

[0066] Perform crosshairs image compensation (for missing ): read the source tile of the left channel at the same display order number and same grid position . Apply a leftward translation according to the rule, with a translation amount of due to the right channel being missing. Generate temporary preview tile pixel coordinates from the pixel coordinates of the source tile . Discard out-of-bounds pixels, fill in the gaps with zero values. Then create a seam of width on the shared boundary with adjacent tiles, and perform channel average transition on the seam pixels. For example, at a certain seam pixel, the red, green, blue, and alpha values of the source tile and the adjoining tile are and , respectively, then the written-back value is , where denotes rounding to the nearest integer. The generated temporary preview tile is marked as crosshairs source, waiting for the subsequent original source tile to overwrite.

[0067] The p = 2 and p = 3 positions do not add new tiles, and reuse the last retained tile set of p = 1. The display order numbers are , respectively.

[0068] p = 4 (forward-looking marker, prefer quaternary): according to the direction of the last two touch displacements, the moving direction is calculated to be right-down. A queue of quaternary forward-looking tiles is established, with a selection order of 4 tiles in the right-down direction. Each channel writes the head 4 quaternary tiles in p = 4 to the corresponding position in the display buffer, with a level marker , a forward-looking source field of 1, and a display order number . If the unique number has not arrived continuously within the two time windows corresponding to and , the packet loss count reaches 2, triggering a downshift request: the target view level is reduced from tertiary to quaternary, requesting (the corresponding position in the quaternary grid mapped by scaling). After receiving this quaternary tile in the subsequent time window , directly overwrite the temporary content in the corresponding position of the right channel.

[0069] Statistical the rendering time within this loop. If the rendering time measured at is greater than , then the target view level is uniformly downgraded by one level (original level -> sub-level, sub-level -> tertiary level, tertiary level -> quaternary level) at the 1st reserved marker of the next duty cycle. In this example, the target view level is quaternary after the downgrade, which ensures the priority of binocular synchronization and fluency. When the dragging intensity drops to the low threshold and lasts for 120 milliseconds, the stabilization is triggered; the current preview duty cycle is frozen, and the binocular lockstep marker remains unchanged. The tertiary viewport is mapped back to the original level with a mapping ratio of . .

[0070] . .

[0071] The column and row ranges covered by the original level viewport are , . The original level target set is the direct product of row and column , which contains tiles. The priority tile table is constructed with the geometric center of the original level viewport as the starting point. The geometric center is .

[0072] , and the row and column are . Therefore, the first priority tile is . The priority tile table generates a sequence by expanding outward in each ring and recording the unique number in a clockwise direction until 28 tiles are covered; the left and right channels generate corresponding unique number lists, for example, the first unique number of the left channel is .

[0073] In each display order number time window, the original level tiles of the left and right channels are pulled in pairs and written into the current view buffer in the order of the priority tile table; if there is any low-level content (quaternary front tile, tertiary reserved marker tile, or temporary preview tile from the cross-eye source) at the target position, it is covered by the original source tile. If a unique number does not arrive in two consecutive time windows, the original level refill request is repeated for that unique number until it is successfully written. For all temporary preview tiles marked as cross-eye sources, the corresponding original source unique number is read; once the original source tile arrives and its fixed length fingerprint matches the index, it is covered and the cross-eye source metadata is cleared. When all 28 original level target tiles are written in both channels, it is determined that the current viewport area has reached the original level complete display; the frozen state is released, the binocular lockstep marker and lockstep display time window values remain unchanged, and a new browsing action is awaited.

[0074] , for example Figure 2As shown, the tile transmission and packet loss rate analysis curve of the present application demonstrates the dynamic change process of tile transmission performance in the binocular 4K picture fast browsing method. The control parameters set in the chart include a lockstep display time window of 16.7 ms, a continuous non-arrival threshold of 2 times, a level fallback trigger condition of continuous packet loss, a forward-looking supplement delay of 5 ms, and a maximum retransmission number of 3 times. The abscissa in the figure represents the time window number, and the ordinate left represents the transmission success rate percentage, and the right corresponds to the packet loss rate alarm index. The left eye channel tile transmission success rate curve is represented by a blue solid line, and the right eye channel tile transmission success rate curve is represented by an orange solid line. During time windows 1-2, the transmission success rates of the two channels are maintained at a relatively high level of more than 85%. However, starting from time window 3, the transmission success rate shows a significant downward trend, and the left eye channel drops to about 40% in time window 4, and the right eye channel drops to about 35% at the same period, both of which are lower than the baseline of 85% of the transmission success rate and the threshold line of 30% of the packet loss rate alarm. The cross-eye image compensation trigger times marked by the purple rectangular area in the figure occur in time windows 3, 4, and 5, respectively, among which the compensation strength in time window 4 is the highest. Through the joint action of the forward-looking supplement and level fallback mechanisms, the transmission success rate gradually rises from time window 5, and stabilizes at a level of more than 85% during time windows 6-8. The key performance indicators on the right side of the chart show that the average transmission success rate reaches 82.5%, the cross-eye compensation trigger is 3 times, the forward-looking supplement success rate is 95%, and the level fallback recovery time is 150 ms, confirming the adaptive compensation capability of the method of the present application under abnormal network transmission conditions.

[0075] As Figure 3As shown, the duty cycle sequence switching and transmission efficiency analysis curve details the optimization effect of the preview duty cycle sequence dynamic switching mechanism in the present application on transmission efficiency. The analysis is based on the following control parameters: set skip threshold of 70%, set reserved threshold of 60%, duty cycle sequence length of 10 positions, switching judgment period of 5 cycles, and transmission bandwidth limit of 100 MB / s. The abscissa in the figure represents the duty cycle sequence number, the left ordinate represents the transmission efficiency percentage, and the right ordinate represents the duty cycle percentage. The main transmission efficiency curve is drawn in blue solid line, the skip mark duty cycle is represented in red dotted line, and the reserved mark duty cycle is represented in green dotted line. During duty cycles 1-3, the system is in fast browsing state, as indicated by the red area in the figure, at this time the skip mark ratio exceeds the set threshold of 70%, but the transmission efficiency is relatively low, only maintained in the range of 20-45%. The key sequence switching events occur in cycles 4 and 5, marked as sequence switching point 1 and sequence switching point 2, respectively. Through intelligent switching of the duty cycle sequence template, the system is converted from fast browsing state to ordinary browsing state, as shown in the green area in the figure, and the reserved mark ratio is increased to more than 60%. This switching results in significant improvement in transmission efficiency, which has been rising since cycle 5 and finally stabilized at a high efficiency level of 80-85% during cycles 8-10. The sequence switching effect analysis data shows that the average response time of 2 sequence switching is 50 ms, the transmission efficiency improvement amplitude reaches 50%, the bandwidth utilization rate is improved by 40%, the user experience is improved by 65%, and the sequence switching success rate reaches 100%, fully verifying the technical advantages of the duty cycle sequence adaptive switching mechanism of the present application.

[0076] Although the specific embodiments of the present application are described above, those skilled in the art should understand that these specific embodiments are only illustrative, and those skilled in the art can make various omissions, substitutions and changes to the details of the above-mentioned methods and systems without departing from the principles and essence of the present application. For example, combining the above-mentioned method steps, so as to perform substantially the same function in substantially the same manner to achieve substantially the same result, is within the scope of the present application. Therefore, the scope of the present application is only limited by the appended claims.

Claims

1. A binocular 4K image fast browsing method, characterized in that: The method includes: Step 1: Read the touch coordinates and touch time from the system touch callback, cache at least three consecutive touch callbacks in chronological order, calculate the drag strength corresponding to the displacement and interval between two adjacent touch callbacks, and if the drag strength reaches the high level threshold for two consecutive times, enter the fast browsing state; otherwise, maintain the normal browsing state; Step 2: Based on the browsing state determination results, a hierarchical view index is constructed and a tile fingerprint chain covering tiles at all levels is generated. A preview duty cycle consisting of a retain flag, a skip flag, and a forward-looking flag is generated based on the browsing state. This preview duty cycle includes a binocular lock-step flag for synchronizing the left and right eye channels. Based on the duty cycle position and the binocular lock-step flag, a set of tiles related to the view level and viewport are synchronously selected. The selected tiles are rearranged, losslessly encapsulated, and transmitted sequentially. The decoder unpacks and renders tiles from the left and right eye channels in pairs within the lock-step display time window. If missing tiles are encountered, temporary preview tiles are generated by cross-eye mirror compensation. The duty cycle continues to advance, and forward-looking replay and level fallback operations are executed when triggered by delay or packet loss. Step 3: When it is detected that the drag intensity drops to a low level threshold and lasts for a preset time, or when a touch end event is detected, the current preview duty cycle sequence is frozen and a steady replacement and gradual recovery process is entered.

2. The binocular 4K image fast browsing method according to claim 1, characterized in that: In step 3, the process of steady-state replacement and gradual recovery is as follows: backfill the original-level tiles from the center outward in a ring-by-ring manner according to the priority tile table, replace the temporary preview tiles marked as cross-eye sources with the original source tiles one by one, and keep the binocular lock-step mark and lock-step display time window unchanged during the entire replacement period until the current viewport area reaches the original-level complete display.

3. The binocular 4K image fast browsing method according to claim 2, characterized in that: In step 1, a first displacement distance and a first time interval are calculated based on the first touch record and the second touch record; a second displacement distance and a second time interval are calculated based on the second touch record and the third touch record; wherein, the displacement distance is calculated according to the pixel distance between two points in the screen pixel coordinate system of the two touch coordinates; the time interval is calculated according to the millisecond difference between the two touch times; the first displacement distance is ratioed to the first time interval to obtain the first drag strength; the second displacement distance is ratioed to the second time interval to obtain the second drag strength; when any time interval is less than the preset effective time interval threshold, the current judgment is discarded and the effective time interval is completed by waiting for the next system touch callback.

4. The binocular 4K image fast browsing method according to claim 3, characterized in that: For each image of the left and right eyes, a hierarchical view index of four levels, namely original, secondary, tertiary and quaternary, is established in the memory. Each level is composed of a grid of tiles with fixed side lengths, and each tile has a unique number and a fixed-length fingerprint; the fixed-length fingerprint is generated by the pixel content of the tile through an irreversible summary function.

5. The binocular 4K image fast browsing method according to claim 4, characterized in that: In step 2, the fixed-length fingerprints of each tile are connected in the order of tile numbers to form a tile fingerprint chain.

6. The binocular 4K image fast browsing method according to claim 5, characterized in that: In step 2, the process of generating a preview duty sequence based on the browsing state includes: selecting a preview duty sequence template based on the browsing state judgment result, and the duty sequence template is arranged in a cyclic order by three types of marks: retention mark, skip mark and look-ahead mark; when in the fast browsing state, a duty sequence template with a skip mark ratio exceeding the set skip threshold is selected; when in the normal browsing state, a duty sequence template with a retention mark ratio exceeding the set retention threshold is selected; each duty sequence template contains a binocular lock-step mark to ensure that the left and right eye channels execute a consistent duty relationship in the same cycle.

7. The binocular 4K image fast browsing method according to claim 6, characterized in that: In step 2, the tile set selected from the lock-step selection is rearranged and losslessly encapsulated based on the tile fingerprint chain; each encapsulated fragment contains a level tag header, the number of tiles, a display order number, and a tile data area. The level tag header of the encapsulated fragment indicates the view level and whether the encapsulated fragment comes from the front-view tag. The display order number is used by the decoding end to restore the display order according to the occupied position.

8. The binocular 4K image fast browsing method according to claim 7, characterized in that: In step 2, the decoder pairs the encapsulated fragments of the left and right eye channels according to the display order number, and takes out the fragments of the same occupancy position in pairs within the lock-step display time window for rendering; if a channel lacks a fragment in the current time window, it enters the cross-eye mirror compensation process; when a channel lacks the target tile at the current occupancy position, obtains the tile at the same occupancy position, the same view level, and the same tile number from the other channel to generate a temporary preview tile; the generation method is: copy the obtained tile, perform geometric translation according to the preset fixed lateral offset, perform boundary cropping and pixel-level stitching with the adjacent tiles; the temporary preview tile is marked as a cross-eye source and can be covered by subsequent original source tiles; at the occupancy position corresponding to the front-view mark, the front-view tile from the fourth-level view is rendered first to reduce visible delay; if the decoder detects that the same tile has not arrived in two consecutive time windows, the view level of the tile is reduced by one level and re-requested until it is filled in the subsequent time window or the stop-and-go replacement process of step 3 is entered.

9. The binocular 4K image fast browsing method according to claim 8, characterized in that: In step 2, after completing a duty cycle, decide whether to continue using the same preview duty sequence or switch to a template with a higher retention mark ratio based on the current browsing status; when the rendering time exceeds the lock-step display time window, immediately retract the view level of the next cycle by one level.

Images