WO2017047494A1 - 画像処理装置 - Google Patents
画像処理装置 Download PDFInfo
- Publication number
- WO2017047494A1 WO2017047494A1 PCT/JP2016/076445 JP2016076445W WO2017047494A1 WO 2017047494 A1 WO2017047494 A1 WO 2017047494A1 JP 2016076445 W JP2016076445 W JP 2016076445W WO 2017047494 A1 WO2017047494 A1 WO 2017047494A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- disturbance
- correction
- disturbances
- processing apparatus
- Prior art date
Links
- 238000012545 processing Methods 0.000 title claims abstract description 91
- 238000001514 detection method Methods 0.000 claims abstract description 99
- 238000012937 correction Methods 0.000 claims abstract description 76
- 238000000034 method Methods 0.000 claims abstract description 63
- 230000008569 process Effects 0.000 claims abstract description 33
- 230000000694 effects Effects 0.000 claims abstract description 20
- 238000012706 support-vector machine Methods 0.000 claims description 16
- 230000006872 improvement Effects 0.000 claims description 4
- 238000003702 image correction Methods 0.000 abstract description 29
- 239000003595 mist Substances 0.000 abstract 1
- 238000004458 analytical method Methods 0.000 description 39
- 230000037303 wrinkles Effects 0.000 description 34
- 230000006870 function Effects 0.000 description 20
- 238000003384 imaging method Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 15
- 238000004364 calculation method Methods 0.000 description 12
- 238000000605 extraction Methods 0.000 description 8
- 239000004071 soot Substances 0.000 description 8
- 238000010191 image analysis Methods 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000003707 image sharpening Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 230000002411 adverse Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000004020 conductor Substances 0.000 description 2
- 230000006866 deterioration Effects 0.000 description 2
- 210000000744 eyelid Anatomy 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/254—Analysis of motion involving subtraction of images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/60—Noise processing, e.g. detecting, correcting, reducing or removing noise
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/44—Event detection
Definitions
- the present invention relates to an image processing apparatus.
- the automatic detection function include a technique for detecting an abnormal operation of a subject in an image, and a technique for detecting only a specific person shown on a surveillance camera.
- a wrinkle correction technique for correcting the degradation of the image and techniques for improving the image quality.
- a flame correction technique for improving the image quality.
- an image sharpening technique for improving the resolution and resolution of an image by adaptive edge enhancement processing.
- the wrinkle correction technique may emphasize noise components by contrast enhancement.
- the flame correction technique may cause blurring in an area where a moving object exists.
- the image sharpening technique sometimes emphasizes a noise component for an image having a large noise component.
- Patent Document 2 discloses an invention that stably detects an object by performing gradation correction on an image in which the contrast of the image is lowered entirely or partially. Yes.
- Patent Documents 3 and 4 disclose inventions for correcting the heat based on time smoothing.
- An image processing apparatus of the present invention is an image processing apparatus having a disturbance detection unit, an image correction unit, and an event detection unit, and the image correction unit switches an image correction method based on disturbance information output from the disturbance detection unit.
- the disturbance information is, for example, a disturbance level.
- the residual disturbance or its influence on event detection is evaluated.
- an optimal correction method is automatically determined that is an acceptable residual disturbance or influence and can be processed by given hardware.
- Disturbances include, for example, noise reduction caused by low contrast due to glaze, distortion of the subject due to heat, and gain increase under low illuminance. This may include effects that the camera is indirectly affected by.
- the influence of wrinkles can be estimated from the contrast bias for each block, which is a screen divided into a plurality of blocks.
- the influence of the hot flame can be estimated by comparing the difference information of the histogram for each block and the difference information of the pixel value.
- the influence of noise can be estimated from the parameters of the imaging unit.
- FIG. 1 is a block diagram of an image processing apparatus 1 according to one embodiment.
- FIG. 3 is a block diagram of a disturbance detection unit 102 of the image processing device 1. The figure for demonstrating the image analyzed by the disturbance detection part 102. The flowchart of the acquisition of the analysis image by the analysis image extraction part 201.
- FIG. 3 is a block diagram of the eyelid influence detection unit 202 of the image processing device 1. The block diagram of the influence detection part ⁇ 203 of the positive flame of the image processing apparatus 1.
- 10 is a flowchart of a method for determining a disturbance level of the image processing apparatus IV.
- FIG. 3 is a block diagram of an event detection unit 104 of the image processing device 1. The figure explaining the effect
- FIG. 10 is a block diagram of an image processing apparatus 101 according to a second embodiment. The block diagram of the event detection part 134 of the image processing apparatus 101.
- FIG. 1 is a block diagram of an image processing apparatus 1 according to an embodiment.
- the image processing apparatus 1 includes a disturbance detection unit 102, an image correction unit 103, and an event detection unit 104.
- the image processing apparatus 1 detects and outputs an event from an input image (image signal) 111 from the imaging unit 101.
- the image processing apparatus 1 has a part of the functions in the imaging unit 101 to perform processing in a distributed manner, or processes input images from the plurality of imaging units 101 in a centralized manner with one integrated server or the like. It can also be realized in a manner.
- the imaging unit 101 captures a subject as a moving image and outputs it as an input image 111 of the disturbance detection unit 102 and the image correction unit 103. Also, the setting parameter 112 at the time of shooting is output to the disturbance detection unit 102.
- the disturbance detection unit 102 analyzes the input image 111, detects the influence of the disturbance received by the input image 111, and outputs it as disturbance information 113 to the image correction unit 103 and the event detection unit 104.
- the disturbance includes, for example, soot, heat and noise.
- the setting parameter 112 of the imaging unit 101 may be used.
- the output disturbance information 113 represents, for example, the intensity of disturbance (the degree of deterioration of the input image 111) in three levels (high, medium, low) and the like.
- the image correction unit 103 performs a correction process on the input video 111 according to the disturbance information 113 and outputs the corrected image 114 and the correction information 115 to the event detection unit 104.
- a process for reducing the influence of disturbances such as a wrinkle correction and a flame correction, and a technique for improving image quality such as sharpening (super-resolution) are used.
- the processing amount of the necessary image correction processing exceeds the allowable processing amount for real-time processing, the processing is prioritized based on the disturbance information 113, and only the one with higher priority is processed. For example, the priority is determined such that the higher the disturbance, the higher the priority.
- the correction information 115 is information indicating the correction processing actually performed on the input video 111, and includes not only the presence / absence of correction but also information such as the intensity of the correction or the degree of improvement, the applied area range, and the like. sell.
- the event detection unit 104 detects an event from the corrected image 114 according to the disturbance information 113 and the correction information 115. First, the event detection unit 104 estimates the influence of the disturbance remaining in the corrected image from the disturbance information 113 and the correction information 115. Depending on the estimated result, the detection process or those parameters are adaptively switched.
- FIG. 2 is a block diagram of the disturbance detection unit 102 of the image processing apparatus 1 of this example.
- the disturbance detection unit 102 includes an analysis image extraction unit 201, a soot influence detection unit 202, a hot flame effect detection unit 203, and a disturbance level determination unit 204.
- the analysis image extraction unit 201 extracts an image block for detecting the influence of soot and heat from the input image 111, and outputs the image block as a soot analysis target image 211 and a heat flame analysis target image 212.
- the disturbance level determination unit 204 detects the influence of noise caused by the decrease in contrast due to haze, distortion of the subject due to haze, and gain increase under low illumination, based on the input haze detection result 213, haze detection result 214, and setting parameter 112. Determine and output as disturbance information 113.
- the influence of noise on the input image can be estimated by the setting parameter 112 of the imaging unit 101.
- the SN ratio (Signal to Noise ratio) of the imaging unit 101 depends on the performance of the imaging sensor itself, and is calculated as follows using measurable parameters according to NMVA1288.
- ⁇ p is the number of incident photons
- ⁇ is quantum efficiency
- ⁇ d dark noise (thermal noise ⁇ d c and readout noise ⁇ d 0)
- ⁇ o is spatial offset noise
- S g is spatial gain noise.
- ⁇ d and ⁇ o are dominant in the denominator.
- SNR ⁇ ( ⁇ p ) 1/2 is relatively large, so SNR ⁇ ( ⁇ p ) 1/2 . This is a region where so-called light shot noise dominates. At higher illumination, it saturates SNR ⁇ 1 / S g.
- ⁇ p or SNR it is difficult to accurately obtain ⁇ p or SNR from the brightness of the input image 111 that has undergone AGC or the like.
- AGC analog AGC control value
- the SNR can be basically described as a monotonically decreasing function with respect to the AGC gain, even when AE (automatic exposure) control is taken into consideration.
- FIG. 3 is a diagram for explaining an image analyzed by the disturbance detection unit 102.
- the input image 111 is equally divided into a plurality of blocks.
- FIG. 3A shows that one of the blocks (upper left corner) is used for the analysis target image (time t-1) at time t-1.
- FIG. 3B shows that a block different from the block used at time t ⁇ 1 (for example, right adjacent) is used for the analysis target image at time t (time t).
- FIG. 3C shows that another block (for example, right adjacent) is used for the analysis target image (time t + 1) at time t + 1.
- the 2 may divide the input image 111 into blocks and extract one of these blocks as an analysis target image by sequential scanning, for example, as shown in FIG. 3.
- the entire input image is analyzed by taking a plurality of frames, there is no problem in the occurrence of the disturbance because the temporal change is usually slow.
- FIG. 4 is a flowchart of the analysis image acquisition by the analysis image extraction unit 201.
- the analysis image extraction unit 201 starts processing (Start), acquires the input image 111 (S401), divides the input image 111 into a predetermined number of blocks (S402), and extracts the analysis target block. (S403), a comparison block stored in an image memory (not shown) is acquired (read) (S404), and the analysis target block and the past block (comparison block) are output as analysis images (S405).
- a block at the same position as the analysis target block of the next input image 111 is stored (written) in an image memory (not shown) as a comparison block (S406), and the process ends (End).
- the analysis image extraction unit 201 may store an image block at the same position as the position of the image block to be analyzed in the next input frame, for example, in an image memory (not shown).
- FIG. 5 is a block diagram of the eyelid influence detection unit 202 of the image processing apparatus 1.
- the wrinkle influence detection unit 202 detects the influence of wrinkles from the contrast deviation of the wrinkle analysis target image 211 given from the analysis image extraction unit 201, and outputs the influence as a wrinkle detection result 213.
- the wrinkle effect detection unit 202 includes a wrinkle correction processing unit 301, an image comparison unit 302, and a wrinkle effect calculation unit 303.
- the wrinkle correction processing unit 301 performs wrinkle correction processing on the wrinkle analysis target image 211 and outputs it as a wrinkle correction image 311.
- the wrinkle correction processing is generally composed of gradation correction and spatial filter processing.
- the image comparison unit 302 compares the wrinkle analysis target image 211 and the wrinkle correction image 312 and outputs the difference information 312.
- the difference information 312 may be a processing result such as SAD (Sum of Absolute Difference).
- the wrinkle influence calculation unit 303 calculates the wrinkle detection result 213 from the input difference information 312 and outputs it.
- the fact that the difference information 312 has a large value means that the wrinkle correction processing has been applied strongly, and represents that the wrinkle analysis target image 211 has been deteriorated due to a decrease in contrast.
- the wrinkle detection result 213 is not limited to the one obtained above, but is a bias or statistical variance of the pixel luminance value of the wrinkle analysis target image 211 on the histogram, or calculated based on spatial frequency spectrum analysis. May be.
- FIG. 6 is a block diagram of the heat effect detection unit 203 of the image processing apparatus 1.
- the heat effect detection unit 203 detects the influence of the heat by detecting fluctuations in the background area where no moving object exists from the input heat flame analysis target image 212 to detect the heat. Output as result 214.
- the hot flame influence detection unit 203 includes a moving object detection unit 401, a background region comparison unit 402, and a hot flame influence calculation unit 403 that are robust against fluctuations.
- the moving object detection unit 401 that is robust to fluctuations detects a moving object from the input heat flame analysis target image 212 and outputs it as moving object region information 411.
- the hot flame analysis target image 212 includes an analysis target image block of the current (time t) input image and an image block at the same position of the previous input image (at time t ⁇ 1).
- the moving object detection unit 401 robust to fluctuations divides the image analysis target image 212 into sub-blocks of 32 ⁇ 32 pixels, and generates histograms h 1 and h 2 quantized with k gradations for each sub-block. To do.
- the following expression is compared between two input sub-blocks, and a region of C> T is determined as a moving object. This uses the characteristic that the influence of fluctuation is robust to the histogram. Even if it puts in the environment with a fluctuation
- the background region comparison unit 402 calculates a difference between regions (sub-blocks) where no moving object exists, out of the two blocks of the image analysis target image 212, and outputs the difference as background region difference information 412. .
- the background area difference information 412 cannot be temporarily used when conductors exist in all areas.
- the flame influence calculation unit 403 calculates the influence of the flame by averaging the input background area difference information 412 and outputs the result as the flame detection result 214. Note that the background area difference information 412 cannot be temporarily used when conductors exist in all areas.
- FIG. 7 is a flowchart for explaining a disturbance level determination method by the disturbance level determination unit 204.
- the procedure shown in FIG. 7 can determine both the heat and flame by appropriately setting the threshold value X for the number of blocks and the threshold value T for the disturbance value.
- the disturbance level determination unit 204 acquires a soot detection result 213 or a flame detection result 214 as a disturbance detection result of the analysis image (S701).
- the disturbance level determination unit 204 holds the disturbance detection results for all the blocks, updates them using the disturbance detection acquired in S701 (S702), and counts the blocks in which the disturbance is detected (S703). If the disturbance detection result cannot be acquired, it is not updated.
- S704 it is determined whether or not the block in which the disturbance is detected is equal to or greater than X. If it is equal to or greater than X (YES), the process proceeds to S705, and if it is less than X (NO), the process proceeds to S709. Proceed to processing.
- the disturbance level determination unit 204 calculates the average disturbance value of the block in which the disturbance is detected (S705), determines whether or not the average disturbance value is T or more (S706), and the average disturbance value is T or more. If (YES), the process proceeds to S707, and if the average disturbance value is less than T (NO), the process proceeds to S708.
- the disturbance level is set to “high”, and the process ends (End).
- the disturbance level is set to “medium” and the process ends (End).
- the disturbance level is set to “low” and the process ends (End).
- the image correction unit 103 in this example has the ability to perform the wrinkle correction, the flame correction, and the sharpening (super-resolution) processing.
- ⁇ correction is not limited to ⁇ , but sharpening (super-resolution) is useful for detecting events from small distant subjects.
- the image correction unit 103 derives the importance of correction from the disturbance level of haze, heat, and noise included in the input disturbance information 113, sorts the importance in order of importance, and allows the processing amount in order of importance. Process in range.
- the image processing apparatus 1 includes, for example, a function g Fog () that calculates the importance of the wrinkle correction process from the disturbance level of the hail, a function g HeatHaze () that calculates the importance of the haze correction process from the disturbance level of the hot flame ,
- the function g Sharpen () which calculates the importance of sharpening (super-resolution) processing from the noise disturbance level, can be used, and the three levels of disturbance levels are also defined as shown in the table below. it can.
- This table is created based on knowledge about which image correction processing is necessary for the current input video image in consideration of each disturbance level.
- each image correction process is prioritized in consideration of the disturbance level and the effect of image correction. That is, even if the adverse effect on event detection is large, disturbances that are hardly improved by correction processing are not dealt with. Since super-resolution processing emphasizes noise, the higher the noise disturbance level, the lower the importance. Further, in order to ensure the real-time property, processing may be performed within the range allowed by the processing amount in descending order of priority.
- FIG. 8 is a block diagram of the event detection unit 104 of the image processing apparatus 1.
- the event detection unit 104 includes a residual disturbance level calculation unit 501, an image analysis unit 502, and a detection processing unit 503.
- the residual disturbance level calculation unit 501 calculates a disturbance level that will remain in the corrected image 114 from the input disturbance information 113 and the correction information 115 and outputs it as the residual disturbance information 511.
- the residual disturbance information 511 may represent the influence of noise such as soot and hot flame in three stages of “high”, “medium”, and “small”, respectively.
- the effect of wrinkles is “small” when wrinkle correction is performed by the image correction unit 103, or a level one level lower than the original level. If the wrinkle correction is not performed, the wrinkle level of the disturbance information 113 may be used as it is.
- the influence of the hot flame is “small” when the wrinkle correction is performed by the image correction unit 103, or the level is lowered by one step from the original level. When the flame correction is not performed, the flame level of the disturbance information 113 may be used as it is.
- the effect of noise is that, when sharpening processing is performed by the image correction unit 103, the noise level of the disturbance information 113 is “low”, and the noise level of the disturbance information 113 is “medium” or “high”.
- the noise level of the disturbance information 113 is output as it is.
- the disturbance information 113 indicates a continuous amount of disturbance level and the correction information 115 indicates a continuous amount of disturbance removal (the degree of improvement by correction)
- the respective residual disturbance levels are also calculated based on the difference between them. can do.
- the image analysis unit 502 outputs the feature amount x obtained by analyzing the corrected image 114 as analysis information 512.
- the detection processing unit 503 detects an event from the analysis information 512 according to the residual disturbance information 511 and outputs a detection result 116.
- the detection processing unit 503 detects an event when the following equation (Equation 2) holds.
- Equation 2 g () is a detection function for detecting a specific event, and it is close to 0 (or becomes smaller as a negative number) so that a specific event is probable, and it is so large that it is not accurate.
- T is a threshold function. If the threshold is increased according to the residual disturbance level, the missed detection ratio can be increased, but the oversight can be reduced. If a threshold value is previously incorporated in the detection function, such as linear discrimination or SVM, the above T () may be added to or subtracted from the threshold value.
- FIG. 9 is a schematic diagram for explaining the operation of the threshold function T () in the detection processing unit 503.
- the feature amount x output from the image analysis unit 502 is distributed in the feature amount space.
- the feature quantities x corresponding to a certain event A are concentrated in a narrow area.
- the feature amount x is considered to be scattered and distributed over a wider area as the disturbance level increases.
- the small region 531 corresponding to the small threshold T () is applied, and the internal feature amount is determined as the event A.
- a medium region 532 is applied when the disturbance level is medium
- a large region 533 is applied when the disturbance level is large.
- the image processing apparatus 1 can perform the main image correction process using a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or an FPGA (Field-Programmable Gate Array).
- the allowable calculation amount of the image processing apparatus 1 is determined from the utilization rate of the CPU or the like. Further, the image processing apparatus 1 can perform detection using an optimum feature amount by adding a so-called learning function that stores processing according to the level of disturbance.
- FIG. 9 is a block diagram showing the configuration of the image correction unit 130 and its periphery according to the second embodiment.
- the imaging unit 131 is different from the imaging unit 101 in that an image correction unit 133 is incorporated.
- the imaging unit 131 outputs shooting parameters such as exposure time, aperture, zoom magnification, optical filter insertion and removal, and setting parameters 112 including AGC gain values.
- the image correction unit 133 sets settings for the imaging unit 131, shooting conditions, and the like. Thus, the captured image is automatically subjected to appropriate image processing and output as a corrected image 114.
- the image correction unit 133 outputs complete correction information 115 describing all correction processes performed on the RAW image read from the image sensor of the imaging unit 131.
- the correction information 115 may be limited to the main correction processing that affects the resolution, the SN ratio, and the like. Backlight correction, fog correction, angle fluctuation correction, heat correction, In addition to resolution processing, image distortion correction such as a fisheye lens can be included.
- the disturbance estimation unit 132 receives the setting parameter 112, the correction image 114, and the correction information 115, estimates the level of disturbance remaining in the correction image 114 using a fully coupled perceptron or a support vector machine (SVM), and stores residual disturbance information 531. Output as.
- the residual disturbance information 531 has a plurality of components, and may be the three components of soot, heat, and noise, as in the residual disturbance information 511, or another reference, for example, luminance error (noise), spatial error, time
- the characteristic amount may be a combination of the four components of the dynamic fluctuation and the color error, or a mixture of them. Further, as in the case of the residual disturbance information 511, it may be classified into three levels or may be a continuous amount.
- the disturbance estimation unit 132 does not need to detect them from the correction image 114, and sets the setting parameter 112 and the correction information 115.
- the perceptron or the like is learned in advance, and online learning may be performed in order to improve the generalization ability with respect to the installation status of the camera and the shooting environment.
- the disturbance level corresponding to each component of the residual disturbance information 531 can be calculated by another method, for example, by means such as the wrinkle influence detection unit 202 from the corrected image 114, or a person can evaluate the corrected image 114. And use them as teacher data.
- the event detection unit 134 receives the corrected image 114 and the residual disturbance information 531, and performs event detection from the corrected image 114 so that the optimum false alarm rate and missed rate considering the residual disturbance information 531 are obtained.
- the event detection unit 134 of this example includes a convolutional neural network (CNN) 151, a support vector machine (SVM) identifier 152, and a risk factor controller 153.
- CNN convolutional neural network
- SVM support vector machine
- the CNN 151 receives the pixel value of the event detection unit 134, repeats the local convolution layer and the maximum value pooling layer a plurality of times while reducing the number of data, and if necessary, passes through all the coupling layers to obtain a plurality of values ( (Feature vector) is output.
- This feature vector corresponds to the analysis information 512 of the previous embodiment.
- the correction information 115 may be input to a layer in the middle of the CNN 151 (for example, a total bond layer).
- the SVM discriminator 152 is a soft margin SVM, and outputs a discrimination result (true or false hard decision value) for each specific event to be detected.
- the SVM identifier 152 can be configured using a plurality of 2-class SVMs or 1-class SVMs for each event. In learning, a loss function that penalizes an incorrect solution, such as a hinge function, is used.
- the risk factor controller 153 manages the learning of the SVM discriminator 152 for each combination of a plurality of disturbance levels indicated by the residual disturbance information 531.
- the risk factor controller 153 sets the result of learning the kernel and support vector of the SVM classifier 152 in common regardless of the disturbance level as a representative learning result or a learning result when the disturbance level is the lowest, For other disturbance levels, learning is performed based on the representative learning result and the like. At that time, all samples used for learning are stored, or parameters and function values of loss functions such as C-SVM and ⁇ trick are held.
- the risk factor that the SVM classifier 152 erroneously identifies (more accurately speaking, learns the wrong classification with a large generalization error) Evaluate and tune each disturbance level combination.
- Tuning is a function corresponding to the threshold function T () in the previous embodiment, and emphasizes a low miss rate, so that the risk rate is equal to or less than the desired miss rate. Is called.
- the threshold value b of the discrimination function of the SVM discriminator 152 is adjusted, or parameters such as C, ⁇ , and ⁇ are updated and learned again. Control of the risk factor can be limited so that the false alarm rate does not exceed an acceptable limit.
- the tuning of this example can be performed independently for each event to be detected, and the miss rate and the false alarm rate can be optimized according to the degree of the threat of the event.
- the present invention is widely applied to video processing such as video content analysis that detects annoying or dangerous situations from video captured by a surveillance camera or the like, or extracts a desired event or metadata from TV program material. it can.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Databases & Information Systems (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
Description
また、他の先行技術文献としては、例えば、特許文献2に全体的又は部分的に画像のコントラストが低下した画像に対し階調補正を施すことで安定して物体を検知する発明が開示されている。また、特許文献3及び4に時間平滑化をベースとした陽炎補正の発明が開示されている。
本発明は、前提となるハードウェアを用いて画像からの事象検知をするために、前処理を適切に行うことを目的とする。
なお、外乱には、例えば霞によるコントラストの低下、陽炎による被写体の歪み、低照度下のゲインアップが原因となるノイズの強調などがあげられるが、これらに限定されず、撮影対象や環境の変化に伴いカメラが間接的に受ける影響を含みうる。
検知部は、畳み込みニューラルネットによっても実装することができ、動作中に学習してもよい。
図1は一実施例に係る画像処理装置 1のブロック図である。画像処理装置 1は、外乱検出部 102、画像補正部103、事象検知部 104で構成されている。
画像処理装置 1は、撮像部 101からの入力画像(画像信号) 111から事象を検知して出力する。画像処理装置 1は、例えば、機能の一部を撮像部 101に持たせて分散的に処理したり、複数の撮像部 101からの入力画像を1つの統合サーバ等で集中的に処理したりする様態で実現することもできる。
外乱検出部 102は、入力画像 111を解析し、入力画像 111が受けた外乱の影響を検出し、外乱情報 113として画像補正部 103と事象検知部 104に出力する。
出力される外乱情報 113は、例えば、外乱の強さ(入力画像111の劣化の度合い)を3段階のレベル(高・中・低)等で表すものである。
補正処理には霞補正や陽炎補正などの外乱の影響を軽減する処理や、鮮鋭化(超解像)などの画質の向上を図る技術を使用する。必要な画像補正処理の処理量がリアルタイム化のための許容処理量を超える場合は、外乱情報 113を元に処理に優先順位をつけて、優先順位の高いもののみを処理する。優先順位は、例えば、外乱が強いほど高くなるように決定される。
補正情報 115は、その入力映像111に対して実際に為された補正処理を示す情報であり、補正の有無だけでなく、補正の強度或いは改善の程度、適用された領域範囲等の情報も含みうる。
まず、事象検知部 104は、外乱情報 113と補正情報 115から、補正画像に残留した外乱の影響を推定する。推定した結果に応じて、検知処理もしくはそれらのパラメータを適応的に切り替える。
図2は本例の画像処理装置 1の外乱検出部 102のブロック図である。
図2において、外乱検出部 102は、解析画像抽出部 201と、霞の影響検出部 202と、陽炎の影響検出部 203と、外乱レベル判定部 204で構成されている。
解析画像抽出部 201は、入力画像 111から霞と陽炎の影響を検出するための画像ブロックを抽出し、霞解析対象画像 211と陽炎解析対象画像 212として出力する。
入力画像のノイズの影響は、撮像部 101の設定パラメータ112によって推定できる。撮像部 101のSN比(Signal to Noise ratio)は、撮像センサ自体の性能などに依存し、NMVA1288によれば、測定可能なパラメータを用いて以下の様に計算される。
図3(a)は、そのブロックの一つ(左上隅)を、時刻t-1時の解析対象画像(時刻t-1)に用いることを示している。図3(b)は、時刻t-1に用いたブロックとは別のブロック(例えば右隣り)を、時刻tの解析対象画像(時刻t)に用いることを示している。図3(c)は、更に別のブロック(例えば右隣り)を、時刻t+1時の解析対象画像(時刻t+1)に用いることを示している。
図4において、解析画像抽出部 201は、処理を開始し(Start)、入力画像 111を取得し(S401)、入力画像 111を所定数のブロックに分割し(S402)、解析対象ブロックを抽出し(S403)、図示していない画像メモリ内に記憶している比較用ブロックを取得し(読み出し)(S404)、解析対象ブロックと過去ブロック(比較用ブロック)を解析画像として出力し(S405)、次の入力画像 111の解析対象ブロックと同じ位置のブロックを比較用ブロックとして図示していない画像メモリに保存し(書き込む)(S406)、処理を終了(End)する。
その場合、解析画像抽出部 201は、例えば、図示していない画像メモリ等に、次の入力フレームで解析を行う画像ブロックの位置と同位置の画像ブロックを保存してもよい。
霞の影響検出部 202は、解析画像抽出部 201から与えられた霞解析対象画像 211のコントラストの偏りから霞の影響を検出して、その影響を霞検出結果 213として出力する。
霞の影響検出部 202は、霞補正処理部 301と画像比較部 302と霞影響算出部 303で構成される。
画像比較部 302は、霞解析対象画像 211と霞補正画像 312の比較を行い、その差分情報 312を出力する。差分情報 312は、例えば、SAD(Sum of Absolute Difference)等の処理結果で良い。
霞影響算出部 303は、入力した差分情報 312から霞検出結果 213を算出し出力する。差分情報 312が大きい値を持つということは、霞補正処理が強くかかったことを意味し、霞解析対象画像 211がコントラストの低下による劣化を受けていることを表す。なお、霞検出結果 213は上記で得られるものに限らず、霞解析対象画像 211の画素輝度値のヒストグラム上での偏り或いは統計学的な分散であったり、空間周波数スペクトル分析に基づいて算出してもよい。
陽炎の影響検出部 203は、特許文献3や4の技術と同様に、入力した陽炎解析対象画像 212から動体が存在しない背景領域の揺らぎを検出することにより陽炎の影響を検出して、陽炎検出結果 214として出力する。
陽炎の影響検出部 203は、揺らぎにロバストな動体検出部401と背景領域比較部 402と陽炎影響算出部 403で構成される。
例えば、揺らぎにロバストな動体検出部 401は、陽炎解析対象画像 212を32x32画素のサブブロックに分割し、そのそれぞれのサブブロックに対してk階調で量子化したヒストグラムh1、h2を作成する。二つの入力サブブロック間で下記の式の比較を行い、C>Tの領域を動体として判定する。
陽炎影響算出部 403は、入力された背景領域差分情報 412を平均化するなどし、陽炎の影響を算出し、陽炎検出結果 214として出力する。なお、全ての領域で導体が存在する場合、背景領域差分情報 412は一時的に利用できない。
図7は、外乱レベル判定部204による外乱レベルの決定法を説明するフローチャートである。
図7の手順は、ブロック数の閾値X、外乱値の閾値Tを適切に設定することにより、陽炎と霞のどちらの判定も行うことができる。
外乱レベル判定部 204は、全ブロック分の外乱検出結果を保持しており、S701で取得した外乱検出を用いて更新し(S702)、外乱が検出されたブロックを数える(S703)。なお外乱検出結果が取得できなかったときは更新しない。また、事象検知部 104が行う検知でマスク領域を適用している場合、そのマスク領域に該当する位置のブロックの外乱検出結果は保持する必要が無い。
S704の処理では、外乱が検出されたブロックがX以上であるか否かを判定し、X以上である場合(YES)にはS705の処理に進み、X未満の場合(NO)にはS709の処理に進む。
S708の処理では、外乱レベルを“中”に設定して処理を終了(End)する。
S709の処理では、外乱レベルを“低”に設定して処理を終了(End)する。
本例の画像補正部 103は、上述したように、霞補正、陽炎補正、鮮鋭化(超解像)の処理を行う能力を有しており、少なくとも1種類の処理で有ればリアルタイムでフルフレーム映像について処理できる。霞補正は霞に限らず鮮鋭化(超解像)は遠方の小さな被写体などから事象検知を行う際に有用であり、
画像補正部 103は、入力された外乱情報 113に含まれる霞、陽炎、ノイズの外乱レベルから、補正の重要度を導き出し、重要度が高い順にソートして、重要度が高い順に処理量の許す範囲で処理を行う。画像処理装置 1は、例えば、霞の外乱レベルから霞補正処理の重要度を算出する関数gFog()と、陽炎の外乱レベルから霞補正処理の重要度を算出する関数gHeatHaze()と、ノイズの外乱レベルから鮮鋭化(超解像)処理の重要度を算出する関数gSharpen()と、を用いることができ、3段階の外乱レベルに対しては、以下の表の様にも定義できる。
図8は画像処理装置 1の事象検知部 104のブロック図である。
事象検知部 104は、残留外乱レベル算出部 501と画像解析部 502と検知処理部 503で構成されている。
残留外乱情報 511は、例えば、霞、陽炎等のノイズの影響をそれぞれ“高”、“中”、“小”の3段階で表してもよい。
陽炎の影響は、画像補正部 103で霞補正が行われた場合は“小” 、或いは、元のレベルから1段階下げたレベルにする。陽炎補正が行われなかった場合は、外乱情報 113の陽炎レベルをそのまま用いれば良い。
ノイズの影響は、画像補正部 103で鮮鋭化処理が行われた場合、外乱情報 113のノイズレベルが“小”である場合は“小”、外乱情報 113のノイズレベルが“中”か“高”であった場合は“高”を出力し、画像補正部で鮮鋭化処理が行われていなかった場合、外乱情報 113のノイズレベルをそのまま出力する。
或いは、外乱情報 113が連続量の外乱レベルを示し、補正情報 115が連続量の外乱除去量(補正による改善の程度)を示している場合、それぞれの残留外乱レベルは、それらの差によっても算出することができる。
検知処理部 503は、残留外乱情報 511に応じて解析情報 512から事象の検知を行い、検知結果 116を出力する。例えば、検知処理部 503は下記(式2)の式が成り立つときに事象を検知する。
画像解析部 502が出力する特徴量xは、特徴量空間上に分布する。特に計量空間として最適化されている場合、ある事象Aに該当する特徴量xは、ある狭い領域に密集する。そして、外乱レベルが大きくなるほど、特徴量xは散乱されて、より広い領域に分布すると考えられる。本例では、残留外乱が小さい時に、小さな閾値T()に対応する小さな領域531が適用され、その内部の特徴量が事象Aと判定される。同様に、外乱レベルが中程度のときは中程度の領域532が、外乱レベルが大きいときは大きな領域533が適用される。
また、画像処理装置 1は、外乱のレベルに応じた処理を記憶する、いわゆる学習機能を付加することにより、最適な特徴量を用いた検知が可能となる。
本例では、画像補正部が撮像部の中にあり、かつその動作を自由に制御できない場合を想定して説明する。
図9は第2の実施例に係る画像補正部 130及びその周辺の構成を示すブロック図である。
撮像部 131は、画像補正部 133を内蔵した点で撮像部 101と異なる。撮像部 131は、露光時間、絞り、ズーム倍率、光学フィルタの挿抜等の撮影パラメータやAGCゲイン値を含む設定パラメータ 112を出力する 画像補正部 133は、撮像部 131への設定や、撮影状況などから、撮像画像に自動的に適切な画像処理を施して、補正画像 114として出力する。また画像補正部 133は、撮像部 131の撮像素子から読み出されたRAW画像に対して施された全ての補正処理を記述する完全な補正情報 115を出力する。なお、補正情報 115は、解像度やSN比等に影響を与える主要な補正処理に関するものに絞ってもよく、コントラスト補正の一種である逆光補正や霞霧補正、画角揺れ補正、陽炎補正、超解像処理などの他、魚眼レンズ等の画像歪み補正などが含まれ得る。
画像補正部133が霧霞や陽炎等の必要とされる補正能力を有している場合、外乱推定部 132は、補正画像114からそれらを検出する必要はなく、設定パラメータ 112と補正情報 115を、単純にパーセプトロン等に入力すればよい。
パーセプトロン等は、予め学習されているものとし、更に、カメラの設置状況や撮影環境に対する汎化能力を向上させるために、オンライン学習を行ってもよい。オンライン学習は、残留外乱情報 531の各成分に対応する外乱レベルを、別の方法、例えば補正画像 114から霞の影響検出部 202のような手段で算出したり、人が補正画像 114を評価して与えたりし、それらを教師データに用いて行う。
図10に示されるように、本例の事象検知部 134は、畳み込みニューラルネットワーク(CNN)151と、サポートベクターマシン(SVM)識別器152と、危険率制御器153と、を有する。
111:入力画像、 112:設定パラメータ、 113:画像劣化情報、 114:補正画像、 115:補正情報、 116:解析結果、
201:解析画像抽出部、 202:霧の影響検出部、 203:陽炎の影響検出部、 204:外乱レベル判定部、
211:霧解析対象画像、 212:陽炎解析対象画像、 213:霧検出結果、 214:陽炎検出結果、
301:霧補正処理部、 302:画像比較部、 303:霧影響算出部、 311:霧補正画像、 312:差分情報、
401:揺らぎにロバストな動体検出部、 402:背景領域比較部、 403:陽炎影響算出部、
411:動体領域情報、 412:背景領域差分情報、
501:残留外乱レベル算出部、 502:画像解析部、 503:検知処理部、 511:残留外乱情報、 512:解析情報。
Claims (6)
- 画像から特定の事象を検知する画像処理装置において、
入力画像を解析し、前記入力画像が受けた複数の外乱の影響を検出し、外乱情報として出力する外乱検出器と、
前記外乱情報に応じて前記入力画像に補正処理を施し、補正された画像と実際に施された補正処理を示す補正情報とを出力する画像補正器と、
前記外乱情報と補正情報とから、前記補正された画像に残留する複数の外乱の度合いを推定し、前記外乱の度合いに応じて選択した検知処理を用いて前記特定の事象を検知する事象検知器と、を備えた画像処理装置。 - 請求項1に記載の画像処理装置において、
前記外乱検出器は、前記入力画像が入力されるごとに、位置を変えながら抽出した部分領域の範囲内で、複数の外乱の影響を検出して更新し、前記入力画像の全体が受けている複数の外乱の影響のそれぞれを、多段階に評価することを特徴とする画像処理装置。 - 請求項2に記載の画像処理装置において、
前記画像補正器は、前記多段階に評価されたそれぞれの前記外乱の影響を、補正の重要度に変換し、前記複数の外乱に対応する補正処理を、前記重要度の高い順に処理量の許す範囲で実行するものであり、
前記重要度は、補正を行った場合の改善効果を考慮して定められることを特徴とする画像処理装置。 - 請求項3に記載の画像処理装置において、前記事象検知部は、
前記外乱情報と前記補正情報とから、前記残留する複数の外乱の度合いを算出する残留外乱レベル算出器と、
前記補正された画像から特徴量を抽出する画像解析器と、
前記残留する複数の外乱の度合いの関数であって、特徴量空間上で前記特定の事象が検出されるべき領域を変化させる前記関数を用いて、前記特徴量から前記特定の事象を判定する検知処理器と、を備えたことを特徴とする画像処理装置。 - 前記関数は、前記残留するいずれかの外乱の度合いが大きいほど、前記検出されるべき領域を広げるように作用し、外乱の度合いに関わらず前記検知処理器の見逃し率を小さく維持することを特徴とする画像処理装置。
- 請求項5に記載の画像処理装置において、
前記検知処理器はサポートベクターマシンであり、前記関数は前記サポートベクターマシンの識別関数の中のしきい値を変化させることを特徴とする画像処理装置。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
SG11201801781RA SG11201801781RA (en) | 2015-09-18 | 2016-09-08 | Image-processing device |
US15/757,933 US20180352177A1 (en) | 2015-09-18 | 2016-09-08 | Image-processing device |
JP2017539866A JP6505237B2 (ja) | 2015-09-18 | 2016-09-08 | 画像処理装置 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2015185427 | 2015-09-18 | ||
JP2015-185427 | 2015-09-18 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017047494A1 true WO2017047494A1 (ja) | 2017-03-23 |
Family
ID=58289272
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2016/076445 WO2017047494A1 (ja) | 2015-09-18 | 2016-09-08 | 画像処理装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20180352177A1 (ja) |
JP (1) | JP6505237B2 (ja) |
SG (1) | SG11201801781RA (ja) |
WO (1) | WO2017047494A1 (ja) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019082652A1 (ja) * | 2017-10-25 | 2019-05-02 | 株式会社 東芝 | 画像センサ、人物検知方法、プログラムおよび制御システム |
JP2019212132A (ja) * | 2018-06-06 | 2019-12-12 | キヤノン株式会社 | 画像処理方法、画像処理装置、撮像装置、プログラム、および、記憶媒体 |
JP2020086891A (ja) * | 2018-11-26 | 2020-06-04 | キヤノン株式会社 | 画像処理装置、画像処理システム、撮像装置、画像処理方法、プログラム、および、記憶媒体 |
JP2022033202A (ja) * | 2017-10-25 | 2022-02-28 | 株式会社東芝 | 画像センサ、動体検知方法、プログラムおよび制御システム |
CN115115907A (zh) * | 2022-06-29 | 2022-09-27 | 桂林电子科技大学 | 一种基于cqd蒸馏的低照度目标检测方法 |
JP7406886B2 (ja) | 2019-08-02 | 2023-12-28 | キヤノン株式会社 | 画像処理装置、画像処理方法、およびプログラム |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107872608B (zh) * | 2016-09-26 | 2021-01-12 | 华为技术有限公司 | 图像采集设备及图像处理方法 |
JP2020046774A (ja) * | 2018-09-14 | 2020-03-26 | 株式会社東芝 | 信号処理装置、距離計測装置、および距離計測方法 |
JP7292123B2 (ja) * | 2019-06-20 | 2023-06-16 | キヤノン株式会社 | 撮像装置及びその制御方法、プログラム、記憶媒体 |
CN113553937A (zh) * | 2021-07-19 | 2021-10-26 | 北京百度网讯科技有限公司 | 目标检测方法、装置、电子设备以及存储介质 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003346160A (ja) * | 2002-05-29 | 2003-12-05 | Matsushita Electric Ind Co Ltd | 移動物体検出装置および移動物体検出方法 |
JP2012084009A (ja) * | 2010-10-13 | 2012-04-26 | Secom Co Ltd | 画像センサ |
JP2013186635A (ja) * | 2012-03-07 | 2013-09-19 | Hitachi Kokusai Electric Inc | 物体検知装置、物体検知方法及びプログラム |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5374220B2 (ja) * | 2009-04-23 | 2013-12-25 | キヤノン株式会社 | 動きベクトル検出装置およびその制御方法、ならびに撮像装置 |
US9451165B2 (en) * | 2013-02-14 | 2016-09-20 | Hitachi Kokusai Electric Inc. | Image processing apparatus |
JP5908174B2 (ja) * | 2013-07-09 | 2016-04-26 | 株式会社日立国際電気 | 画像処理装置及び画像処理方法 |
-
2016
- 2016-09-08 JP JP2017539866A patent/JP6505237B2/ja active Active
- 2016-09-08 US US15/757,933 patent/US20180352177A1/en not_active Abandoned
- 2016-09-08 WO PCT/JP2016/076445 patent/WO2017047494A1/ja active Application Filing
- 2016-09-08 SG SG11201801781RA patent/SG11201801781RA/en unknown
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003346160A (ja) * | 2002-05-29 | 2003-12-05 | Matsushita Electric Ind Co Ltd | 移動物体検出装置および移動物体検出方法 |
JP2012084009A (ja) * | 2010-10-13 | 2012-04-26 | Secom Co Ltd | 画像センサ |
JP2013186635A (ja) * | 2012-03-07 | 2013-09-19 | Hitachi Kokusai Electric Inc | 物体検知装置、物体検知方法及びプログラム |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019082652A1 (ja) * | 2017-10-25 | 2019-05-02 | 株式会社 東芝 | 画像センサ、人物検知方法、プログラムおよび制御システム |
JP2019080177A (ja) * | 2017-10-25 | 2019-05-23 | 株式会社東芝 | 画像センサ、人物検知方法、プログラムおよび制御システム |
JP7002912B2 (ja) | 2017-10-25 | 2022-01-20 | 株式会社東芝 | 画像センサ、人物検知方法、プログラムおよび制御システム |
JP2022033202A (ja) * | 2017-10-25 | 2022-02-28 | 株式会社東芝 | 画像センサ、動体検知方法、プログラムおよび制御システム |
JP7286747B2 (ja) | 2017-10-25 | 2023-06-05 | 株式会社東芝 | 画像センサ、動体検知方法、プログラムおよび制御システム |
JP2019212132A (ja) * | 2018-06-06 | 2019-12-12 | キヤノン株式会社 | 画像処理方法、画像処理装置、撮像装置、プログラム、および、記憶媒体 |
JP7146461B2 (ja) | 2018-06-06 | 2022-10-04 | キヤノン株式会社 | 画像処理方法、画像処理装置、撮像装置、プログラム、および、記憶媒体 |
JP2020086891A (ja) * | 2018-11-26 | 2020-06-04 | キヤノン株式会社 | 画像処理装置、画像処理システム、撮像装置、画像処理方法、プログラム、および、記憶媒体 |
JP7246900B2 (ja) | 2018-11-26 | 2023-03-28 | キヤノン株式会社 | 画像処理装置、画像処理システム、撮像装置、画像処理方法、プログラム、および、記憶媒体 |
JP7406886B2 (ja) | 2019-08-02 | 2023-12-28 | キヤノン株式会社 | 画像処理装置、画像処理方法、およびプログラム |
CN115115907A (zh) * | 2022-06-29 | 2022-09-27 | 桂林电子科技大学 | 一种基于cqd蒸馏的低照度目标检测方法 |
CN115115907B (zh) * | 2022-06-29 | 2024-03-29 | 桂林电子科技大学 | 一种基于cqd蒸馏的低照度目标检测方法 |
Also Published As
Publication number | Publication date |
---|---|
US20180352177A1 (en) | 2018-12-06 |
JPWO2017047494A1 (ja) | 2018-07-12 |
JP6505237B2 (ja) | 2019-04-24 |
SG11201801781RA (en) | 2018-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2017047494A1 (ja) | 画像処理装置 | |
JP5908174B2 (ja) | 画像処理装置及び画像処理方法 | |
CN108335279B (zh) | 图像融合和hdr成像 | |
Kim et al. | Optimized contrast enhancement for real-time image and video dehazing | |
KR100791391B1 (ko) | 노이즈 저감 방법 및 장치 | |
US8494267B2 (en) | Image processing device, image processing method, and program for normalizing a histogram | |
Kim et al. | A novel approach for denoising and enhancement of extremely low-light video | |
US10614554B2 (en) | Contrast adaptive video denoising system | |
CN107395991B (zh) | 图像合成方法、装置、计算机可读存储介质和计算机设备 | |
TW201445454A (zh) | 提升人臉辨識率的影像處理系統及影像處理方法 | |
TW201310389A (zh) | 使用影像對比增進的移動物件偵測方法 | |
CN111612773B (zh) | 一种红外热像仪及实时自动盲元检测处理方法 | |
KR101336240B1 (ko) | 저장된 영상을 이용한 영상 처리 방법 및 장치 | |
US9355435B2 (en) | Method and system for adaptive pixel replacement | |
US8780235B2 (en) | Image processing method | |
JP2008005365A (ja) | 撮像装置 | |
CN107424134B (zh) | 图像处理方法、装置、计算机可读存储介质和计算机设备 | |
US11631183B2 (en) | Method and system for motion segmentation | |
CN107292853B (zh) | 图像处理方法、装置、计算机可读存储介质和移动终端 | |
KR20190072643A (ko) | 얼굴 검출 장치 및 그 제어 방법, 그리고 프로그램 | |
KR20190109242A (ko) | 이미지 신호로부터 계단 아티팩트들을 감소시키는 이미지 프로세싱 장치 | |
CN105184747B (zh) | 低照度图像对比度的提升方法 | |
US20130148890A1 (en) | Digital image processing apparatus and method | |
JP2018160024A (ja) | 画像処理装置、画像処理方法及びプログラム | |
CN113870300A (zh) | 图像处理方法、装置、电子设备及可读存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16846365 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2017539866 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11201801781R Country of ref document: SG |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 16846365 Country of ref document: EP Kind code of ref document: A1 |