# US 10,375,407 B2 · Adaptive thresholding for computer vision on low bitrate compressed video streams > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,375,407 B2 - Document ID: US10375407B2 - Status: Granted - Kind: B2 - Application number: 15/888,927 - Filed: 2018-02-05 - Published: 2019-02-07 - Granted: 2019-08-06 - Earliest claimed: 2018-02-05 - Assignees: Intel Corporation - Applicants: Intel Corporation - Inventors: Srenivas Varadarajan; Omesh Tickoo; Vallabhajosyula Somayazulu; Yiting Liao; Ibrahima Ndiour; Shao-Wen Yang; Yen-Kuang Chen - Web page: https://patentgrep.com/patent/US10375407B2 ## CPC - H04N 19/42: characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation (H04N19/635 takes precedence) - H04N 19/543: using regions - H04N 19/115: Selection of the code volume for a coding unit prior to coding - H04N 19/119: Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks - H04N 19/137: Motion inside a coding unit, e.g. average field, frame or block difference - H04N 19/167: Position within a video image, e.g. region of interest [ROI] - H04N 19/17: the unit being an image region, e.g. an object - H04N 19/176: the region being a block, e.g. a macroblock - H04N 19/513: Processing of motion vectors - H04N 19/70: characterised by syntax aspects related to video coding, e.g. related to compression standards - H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION ## Abstract Techniques related to applying computer vision to decompressed video are discussed. Such techniques may include generating a region of interest in an individual video frame by translating spatial indicators of a first detected computer vision result from a reference video frame to the individual video frame and applying a greater threshold within the region of interest than outside of the region of interest for computer vision evaluation in the individual frame. ## Claims ### Claim 1 A computer-implemented method for applying computer vision to decompressed video comprising: generating a region of interest in an individual video frame by translating one or more spatial indicators of a first detected computer vision result in a reference video frame to the individual video frame; comparing a first computer vision score for a first location of the individual video frame to a first threshold and a second computer vision score for a second location of the individual video frame to a second threshold to determine whether the first score or the second score corresponds to a second detected computer vision result, wherein the first threshold is less than the second threshold in response to the first location being within the region of interest and the second location being outside of the region of interest; and transmitting an indicator of a second detected computer vision result when the first score compares favorably to the first threshold or the second score compares favorably to the second threshold. ### Claim 2 (depends on 1) The method of claim 1, wherein to compare favorably, the first score and the second score must exceed their respective thresholds, and wherein the first threshold is not more than half of the second threshold. ### Claim 3 (depends on 1) The method of claim 1, wherein the first computer vision result comprises an object detection result and the method further comprises: performing object detection on the individual video frame to generate the first and second computer vision scores. ### Claim 4 (depends on 3) The method of claim 3, wherein performing the object detection comprises application of a deep neural network. ### Claim 5 (depends on 1) The method of claim 1, wherein generating the region of interest in the individual video frame comprises: determining one or more coding blocks in the reference video frame that are each overlapped by at least a portion of a bounding region corresponding to the spatial indicators; attaining one or more motion vectors corresponding to the one or more coding blocks; and translating at least the bounding region from the reference video frame to the individual video frame using the one or more motion vectors to generate the region of interest. ### Claim 6 (depends on 5) The method of claim 5, wherein the one or more motion vectors comprise a plurality of motion vectors and translating the bounding region using the one or more motion vectors comprises translating the bounding region using at least one of an average or a median of the plurality of motion vectors. ### Claim 7 (depends on 5) The method of claim 5, further comprising: expanding the bounding region to a boundary defined by the one or more coding blocks to generate the region of interest. ### Claim 8 (depends on 5) The method of claim 5, further comprising: decoding a received bitstream to generate the individual video frame and the reference video frame and to determine the one or motion vectors. ### Claim 9 (depends on 8) The method of claim 8, wherein the individual video frame is a motion predicted frame and the reference video frame is a motion predicted frame coded with fewer bits than the individual video frame. ### Claim 10 (depends on 8) The method of claim 8, wherein the individual video frame is a motion predicted frame and the reference video frame is a stand alone frame. ### Claim 11 (depends on 1) The method of claim 1, further comprising: comparing a third computer vision score for the reference video frame to the first threshold to determine the first detected computer vision result in the reference video frame. ### Claim 12 A system for applying computer vision to decompressed video comprising: a memory to store an individual video frame and a reference video frame; and a processor coupled to the memory, the processor to: generate a region of interest in an individual video frame by translating one or more spatial indicators of a first detected computer vision result in a reference video frame to the individual video frame; compare a first computer vision score for a first location of the individual video frame to a first threshold and a second computer vision score for a second location of the individual video frame to a second threshold to determine whether the first score or the second score corresponds to a second detected computer vision result, wherein the first threshold is less than the second threshold in response to the first location being within the region of interest and the second location being outside of the region of interest; and transmit an indicator of a second detected computer vision result when the first score compares favorably to the first threshold or the second score compares favorably to the second threshold. ### Claim 13 (depends on 12) The system of claim 12, wherein to compare favorably, the first score and the second score must exceed their respective thresholds, and wherein the first threshold is not more than half of the second threshold. ### Claim 14 (depends on 12) The system of claim 12, wherein the first computer vision result comprises an object detection result and the processor is further to: perform object detection on the individual video frame to generate the first and second computer vision scores. ### Claim 15 (depends on 12) The system of claim 12, wherein the processor to generate the region of interest in the individual video frame comprises the processor to: determine one or more coding blocks in the reference video frame that are each overlapped by at least a portion of a bounding region corresponding to the spatial indicators; attain one or more motion vectors corresponding to the one or more coding blocks; and translate at least the bounding region from the reference video frame to the individual video frame using the one or more motion vectors to generate the region of interest. ### Claim 16 (depends on 15) The system of claim 15, wherein the one or more motion vectors comprise a plurality of motion vectors and the processor to translate the bounding region using the one or more motion vectors comprises the processor to translate the bounding region using at least one of an average or a median of the plurality of motion vectors. ### Claim 17 (depends on 15) The system of claim 15, wherein the processor is further to: expand the bounding region to a boundary defined by the one or more coding blocks to generate the region of interest. ### Claim 18 (depends on 12) The system of claim 12, wherein the processor is further to: compare a third computer vision score for the reference video frame to the first threshold to determine the first detected computer vision result in the reference video frame. ### Claim 19 At least one non-transitory machine readable medium comprising a plurality of instructions that, in response to being executed on a device, cause the device to apply computer vision to decompressed video by: generating a region of interest in an individual video frame by translating one or more spatial indicators of a first detected computer vision result in a reference video frame to the individual video frame; comparing a first computer vision score for a first location of the individual video frame to a first threshold and a second computer vision score for a second location of the individual video frame to a second threshold to determine whether the first score or the second score corresponds to a second detected computer vision result, wherein the first threshold is less than the second threshold in response to the first location being within the region of interest and the second location being outside of the region of interest; and transmitting an indicator of a second detected computer vision result when the first score compares favorably to the first threshold or the second score compares favorably to the second threshold. ### Claim 20 (depends on 19) The machine readable medium of claim 19, wherein to compare favorably, the first score and the second score must exceed their respective thresholds, and wherein the first threshold is not more than half of the second threshold. ### Claim 21 (depends on 19) The machine readable medium of claim 19, wherein the first computer vision result comprises an object detection result and the machine readable medium comprises further instructions that, in response to being executed on the device, cause the device to apply computer vision to decompressed video by: performing object detection on the individual video frame to generate the first and second computer vision scores. ### Claim 22 (depends on 19) The machine readable medium of claim 19, wherein generating the region of interest in the individual video frame comprises: determining one or more coding blocks in the reference video frame that are each overlapped by at least a portion of a bounding region corresponding to the spatial indicators; attaining one or more motion vectors corresponding to the one or more coding blocks; and translating at least the bounding region from the reference video frame to the individual video frame using the one or more motion vectors to generate the region of interest. ### Claim 23 (depends on 22) The machine readable medium of claim 22, wherein the one or more motion vectors comprise a plurality of motion vectors and translating the bounding region using the one or more motion vectors comprises translating the bounding region using at least one of an average or a median of the plurality of motion vectors. ### Claim 24 (depends on 22) The machine readable medium of claim 22, wherein the machine readable medium comprises further instructions that, in response to being executed on the device, cause the device to apply computer vision to decompressed video by: expanding the bounding region to a boundary defined by the one or more coding blocks to generate the region of interest. ### Claim 25 (depends on 19) The machine readable medium of claim 19, wherein the machine readable medium comprises further instructions that, in response to being executed on the device, cause the device to apply computer vision to decompressed video by: comparing a third computer vision score for the reference video frame to the first threshold to determine the first detected computer vision result in the reference video frame.