WO2021056808A1 - 图像处理方法及装置、电子设备和存储介质 - Google Patents

图像处理方法及装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2021056808A1
WO2021056808A1 PCT/CN2019/121695 CN2019121695W WO2021056808A1 WO 2021056808 A1 WO2021056808 A1 WO 2021056808A1 CN 2019121695 W CN2019121695 W CN 2019121695W WO 2021056808 A1 WO2021056808 A1 WO 2021056808A1
Authority
WO
WIPO (PCT)
Prior art keywords
iris
feature
image
feature map
processing
Prior art date
Application number
PCT/CN2019/121695
Other languages
English (en)
French (fr)
Inventor
杨凯
徐子豪
费敬敬
吴立威
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Priority to KR1020217008623A priority Critical patent/KR20210047336A/ko
Priority to SG11202013254VA priority patent/SG11202013254VA/en
Priority to JP2021500196A priority patent/JP7089106B2/ja
Priority to US17/137,819 priority patent/US11532180B2/en
Publication of WO2021056808A1 publication Critical patent/WO2021056808A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/18—Eye characteristics, e.g. of the iris
    • G06V40/197—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • 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
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • 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
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/803—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of input or preprocessed data
    • 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
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/18—Eye characteristics, e.g. of the iris
    • G06V40/193—Preprocessing; Feature extraction

Definitions

  • the present disclosure relates to the field of computer vision technology, and in particular to an image processing method and device, electronic equipment, and storage medium.
  • Iris recognition technology uses the lifelong stability and unique characteristics of the iris for identity authentication.
  • the superiority of iris recognition makes it have great application prospects in finance, e-commerce, security, immigration control and other aspects.
  • the present disclosure proposes a technical solution for image processing.
  • an image processing method which includes: acquiring an iris image group, the iris image group including at least two iris images to be compared; detecting the position of the iris in the iris image, and The segmentation result of the iris region in the iris image; performing multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain the iris feature map corresponding to the iris image; using the iris image group The segmentation result and the iris feature map corresponding to the at least two iris images respectively, perform a comparison process, and determine whether the at least two iris images correspond to the same object based on the comparison result of the comparison process .
  • multi-scale feature extraction can be used to extract feature information of multiple scales.
  • the feature information of the bottom and high-level can be obtained at the same time, and then through multi-scale feature fusion, the resulting feature map has higher accuracy and thus more accurate Compare, improve the accuracy of the comparison result.
  • the performing multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain the iris feature map corresponding to the iris image includes: The image region corresponding to the iris position performs the multi-scale feature extraction process to obtain feature maps of multiple scales; using the feature maps of the multiple scales to form at least one feature group, the feature group includes the multiple Feature maps of at least two scales in the scale feature map; based on the attention mechanism, perform the multi-scale feature fusion processing on the feature maps in the feature group to obtain the grouped feature map corresponding to the feature group; based on the feature The grouped feature map corresponding to the grouping is obtained to obtain the iris feature map corresponding to the iris image. Based on the above configuration, the obtained feature maps of multiple scales can be grouped, and the attention mechanism can be further introduced to determine the grouped feature maps of the corresponding groupings, so as to further improve the accuracy of the obtained iris feature maps.
  • the performing the multi-scale feature fusion processing on the feature maps in the feature group based on the attention mechanism to obtain the grouped feature map corresponding to the feature group includes: The connected feature maps of the feature maps of the at least two scales in the grouping perform the first convolution processing to obtain the first sub-feature map; perform the second convolution processing and the activation function processing on the first sub-feature map to obtain The second sub-feature map, the second sub-feature map represents the attention coefficient corresponding to the first sub-feature map; the product result of the first sub-feature map and the second sub-feature map is combined with the first sub-feature map One sub-feature map is added to obtain a third sub-feature map; a third convolution process is performed on the third sub-feature map to obtain a grouped feature map corresponding to the feature group.
  • the obtaining the iris feature map corresponding to the iris image based on the grouping feature map corresponding to the feature grouping includes: performing a weighted sum on the grouping feature map corresponding to each of the grouping features Through processing, the iris feature map corresponding to the iris image is obtained. The grouping characteristics of each grouping are merged by weighted sum to realize the effective fusion of characteristic information.
  • the segmentation result includes a mask image corresponding to an iris region in the iris image, the first identifier in the mask image represents the iris region, and the first identifier in the mask image
  • the second mark indicates a location area outside the iris area.
  • the detecting the position of the iris in the iris image and the segmentation result of the iris area in the iris image includes: performing target detection processing on the iris image, and determining the position of the iris image The position of the iris and the position of the pupil; based on the determined position of the iris and the position of the pupil, the segmentation process is performed on the iris image to obtain a segmentation result of the iris region in the iris image. Based on the above configuration, the detection position corresponding to the iris in the iris image and the segmentation result of the iris area can be accurately determined.
  • the detecting the position of the iris in the iris image and the segmentation result of the iris area in the iris image further includes: respectively determining the image area and the corresponding iris position of the iris image. Performing normalization processing on the segmentation result; performing the multi-scale feature extraction and the multi-scale feature fusion processing on the image region corresponding to the iris position to obtain the iris feature map corresponding to the iris image, and further includes: Performing the multi-scale feature extraction and the multi-scale feature fusion processing on the image region corresponding to the iris position after the normalization process, to obtain an iris feature map corresponding to the iris image. Based on the above configuration, it is possible to perform normalization processing on the image area of the iris position and the segmentation result, and the applicability can be improved.
  • using the segmentation result and the iris feature map respectively corresponding to the at least two iris images to perform the comparison processing includes: using the segmentation results respectively corresponding to the at least two iris images , Determine the first position of the iris region in the at least two iris images; respectively determine the fourth sub-feature map corresponding to the first position in the iris feature map of the at least two iris images; The degree of association between the fourth sub-feature maps respectively corresponding to each iris image determines the comparison result of the at least two iris images.
  • the segmentation results corresponding to different iris images can be used to determine the position of the same iris region in the compared iris images, and the features corresponding to the positions can be used for comparison to obtain the comparison result. Reduce the interference of regional features outside the iris area and improve the accuracy of the comparison.
  • the determining whether the at least two iris images correspond to the same object based on the comparison result includes: an association between the fourth sub-feature maps respectively corresponding to the at least two iris images When the degree is greater than the first threshold, it is determined that the at least two iris images correspond to the same object. Based on the above configuration, through the setting of the first threshold, it can be flexibly adapted to different scenarios, and the comparison result can be conveniently obtained.
  • the determining whether the at least two iris images correspond to the same object based on the comparison result may further include: between the fourth sub-feature maps corresponding to the at least two iris images.
  • the degree of association is less than or equal to the first threshold, it is determined that the at least two iris images correspond to different objects.
  • the image processing method is implemented by a convolutional neural network. Based on the above configuration, the comparison result of the two iris images can be obtained accurately, conveniently and quickly through the neural network.
  • an image processing device which includes: an acquisition module for acquiring an iris image group, the iris image group including at least two iris images to be compared; a detection module for Detect the position of the iris in the iris image and the segmentation result of the iris region in the iris image; a feature processing module for performing multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain The iris feature map corresponding to the iris image; a comparison module, configured to use the segmentation result and the iris feature map corresponding to the at least two iris images to perform a comparison process based on the comparison process The comparison result determines whether the at least two iris images correspond to the same object.
  • the feature processing module is further configured to perform the multi-scale feature extraction process on the image region corresponding to the iris position in the iris image to obtain feature maps of multiple scales; Feature maps of multiple scales to form at least one feature group, the feature group includes feature maps of at least two scales in the multiple scale feature maps; based on the attention mechanism, all feature maps in the feature group are performed
  • the multi-scale feature fusion processing obtains the grouped feature map corresponding to the feature group; and the iris feature map corresponding to the iris image is obtained based on the grouped feature map corresponding to the feature group.
  • the feature processing module is further configured to perform the multi-scale feature extraction process on the image region corresponding to the iris position in the iris image to obtain feature maps of multiple scales; Feature maps of multiple scales to form at least one feature group, the feature group includes feature maps of at least two scales in the multiple scale feature maps; based on the attention mechanism, all feature maps in the feature group are performed
  • the multi-scale feature fusion processing obtains the grouped feature map corresponding to the feature group; and the iris feature map corresponding to the iris image is obtained based on the grouped feature map corresponding to the feature group.
  • the feature processing module is further configured to perform weighting and processing on the grouped feature map corresponding to each of the grouped features to obtain the iris feature map corresponding to the iris image.
  • the segmentation result includes a mask image corresponding to an iris region in the iris image, the first identifier in the mask image represents the iris region, and the first identifier in the mask image
  • the second mark indicates a location area outside the iris area.
  • the second detection module is further configured to perform target detection processing on the iris image, and determine the iris position and the pupil position of the iris image;
  • the segmentation process is performed on the iris image to obtain a segmentation result of the iris region in the iris image.
  • the detection module is further configured to perform normalization processing on the image area corresponding to the iris position of the iris image and the segmentation result respectively;
  • the feature processing module is further configured to: perform the multi-scale feature extraction and the multi-scale feature fusion processing on the image region corresponding to the iris position after the normalization process, to obtain an iris feature map corresponding to the iris image .
  • the comparison module is further configured to use the segmentation results respectively corresponding to the at least two iris images to determine the first position of the at least two iris images that are both iris regions;
  • the comparison result of the at least two iris images is determined according to the degree of association between the fourth sub-feature maps respectively corresponding to the at least two iris images.
  • the comparison module is further configured to determine that the at least two iris images respectively correspond to a fourth sub-feature map with a correlation degree greater than a first threshold.
  • the two iris images correspond to the same object.
  • the comparison module is further configured to determine that the degree of association between the fourth sub-feature maps corresponding to the at least two iris images is less than or equal to a first threshold. At least two iris images correspond to different objects.
  • the device includes a neural network
  • the neural network includes the acquisition module, the detection module, the feature processing module, and the comparison module.
  • an electronic device including:
  • a memory for storing processor executable instructions
  • the processor is configured to call instructions stored in the memory to execute the method described in any one of the first aspect.
  • a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the method described in any one of the first aspect is implemented.
  • a computer program including computer-readable code, when the computer-readable code runs in an electronic device, a processor in the electronic device executes the image Approach.
  • the iris region in the iris image is located and segmented, and the iris position and the iris segmentation result are obtained.
  • multi-scale feature extraction and multi-scale feature extraction can be performed on the iris image. Feature fusion to obtain a high-precision iris feature map, and then use the segmentation result and the iris feature map to perform identity recognition of the iris image to determine whether each iris image corresponds to the same object.
  • the extracted low-level features and high-level features can be fully integrated through multi-scale feature extraction and multi-scale feature fusion, so that the finally obtained iris feature takes into account the texture features of the bottom layer and the classification features of the high level, and improves the accuracy of feature extraction.
  • it can also use the combination of the segmentation result and the iris feature map to consider only the characteristic part of the iris area, reduce the influence of other areas, and more accurately identify whether the iris image corresponds to the same object, and the detection result is higher.
  • Fig. 1 shows a flowchart of an image processing method according to an embodiment of the present disclosure
  • Fig. 2 shows a schematic process diagram of an image processing method according to an embodiment of the present disclosure
  • Fig. 3 shows a flowchart of step S20 in an image processing method according to an embodiment of the present disclosure
  • Fig. 4 shows a schematic diagram of preprocessing of an iris image according to an embodiment of the present disclosure
  • Fig. 5 shows a flowchart of step S30 in an image processing method according to an embodiment of the present disclosure
  • Fig. 6 shows a schematic structural diagram of a neural network according to an image processing method implementing an embodiment of the present disclosure
  • Fig. 7 shows a flowchart of step S33 in an image processing method according to an embodiment of the present disclosure
  • Fig. 8 shows a flowchart of step S40 in an image processing method according to an embodiment of the present disclosure
  • Fig. 9 shows a block diagram of an image processing device according to an embodiment of the present disclosure.
  • FIG. 10 shows a block diagram of an electronic device according to an embodiment of the present disclosure
  • Fig. 11 shows a block diagram of another electronic device according to an embodiment of the present disclosure.
  • the embodiments of the present disclosure provide an image processing method, which can be used to distinguish whether the object corresponding to the iris image is the same object, such as whether it is the iris image of the same person object, based on the iris feature corresponding to the iris image.
  • the execution subject of the image processing method may be an image processing device.
  • the image processing method may be executed by a terminal device or a server or other processing device.
  • the terminal device may be a user equipment (UE), a mobile device, a user terminal, Terminals, cellular phones, cordless phones, personal digital assistants (PDAs), handheld devices, computing devices, in-vehicle devices, wearable devices, etc.
  • the server can be a local server or a cloud server.
  • the image processing method may be implemented by a processor invoking computer-readable instructions stored in the memory.
  • Fig. 1 shows a flowchart of an image processing method according to an embodiment of the present disclosure. As shown in Fig. 1, the image processing method includes:
  • S10 Acquire an iris image group, the iris image group including at least two iris images to be compared;
  • identity verification can be performed through an iris image to identify the identity of an object corresponding to the iris image, or to determine whether the corresponding object has authority.
  • the embodiments of the present disclosure may perform feature processing on the iris image, and realize the comparison of the iris image based on the obtained features, and confirm whether the object corresponding to the iris image is the same object.
  • the corresponding verification operation may be further performed according to whether the determined iris image corresponds to the same object.
  • the embodiments of the present disclosure can first obtain the iris image to be compared, and the iris image to be compared forms an iris image group, and at least two iris images can be obtained.
  • the iris image to be compared in the embodiment of the present disclosure may be collected by an iris camera, or may be transmitted and received by other devices, or may be read from a memory.
  • the foregoing is only an exemplary description. This is not specifically limited.
  • preprocessing may be performed on the iris image first, where the preprocessing may include locating the iris and pupil in the iris image, and determining the positions of the iris and pupil.
  • the iris position and the pupil position can be respectively expressed as the detection frame of the iris and the position corresponding to the detection frame of the pupil.
  • segmentation processing can be further performed on the iris region to obtain corresponding segmentation results, where the segmentation results can be expressed as a mask image.
  • the mask image can be expressed in the form of a vector or a matrix, and the mask image can correspond to the pixels of the iris image one-to-one.
  • the mask image may include a first identifier and a second identifier, where the first identifier indicates that the corresponding pixel is an iris area, and the second identifier indicates that the corresponding pixel is a non-iris area.
  • the first identifier may be "1”
  • the second identifier may be "0", so that the area where the iris is located can be determined by the area formed by the positions of the pixel points of the first identifier in the mask image.
  • S30 Perform multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain an iris feature map corresponding to the iris image;
  • a multi-scale feature extraction process can be performed on the image area corresponding to the iris position. For example, a feature map of at least two scales can be obtained, and then pass Performing convolution processing on the feature map can realize the fusion of features, and then obtain the iris feature map of the iris image.
  • multiple feature maps of different scales of the image region corresponding to the iris position can be obtained in the process of feature extraction.
  • feature extraction can be performed through the residual network, and then the multiple scales of the feature maps can be obtained.
  • the feature map performs convolution processing at least once to obtain an iris feature map that incorporates features of different scales.
  • multi-scale feature extraction the feature information of the bottom and high layers can be obtained at the same time, and the feature information of the bottom and high layers can be effectively fused through multi-scale feature fusion, and the accuracy of the iris feature map can be improved.
  • the attention mechanism can also be used to obtain different attention coefficients for different features.
  • the attention coefficients can indicate the importance of features. Using the attention coefficients to perform feature fusion can be more robust and more efficient. Distinguishing features.
  • S40 Use the segmentation result and the iris feature map respectively corresponding to the at least two iris images to perform a comparison process, and determine whether the at least two iris images correspond to The same object.
  • the segmentation results of at least two iris images that need to be compared can be compared to obtain the positions of both iris regions in the two iris images, based on the features corresponding to the positions of the iris regions
  • the distance between the at least two iris images can be obtained. Wherein, if the distance is less than the first threshold, it indicates that the two iris images compared correspond to the same object, that is, the iris images belonging to the same object. Otherwise, if the distance is greater than or equal to the first threshold, it indicates that the two iris images do not belong to the same object.
  • any two iris images can be compared separately to determine whether any two iris images correspond to the same According to the comparison result of each iris image, the iris image of the same object in the iris image group is determined, and the number of objects corresponding to the iris image in the iris image group can also be counted.
  • FIG. 2 shows a schematic diagram of the process of an image processing method according to an embodiment of the present disclosure, in which two iris images A and B can be acquired first, and pre-processing is performed on the two iris images to obtain the iris position in the iris image , Pupil position and the segmentation results corresponding to the iris area, such as the mask image, and then the iris feature extraction module can be used to perform feature extraction and fusion processing of the image area corresponding to the iris position to obtain the iris feature map of the iris image, and then use the comparison module Based on the mask map and the corresponding iris feature map, the comparison result (score) of whether image A and image B are the same object is obtained.
  • the embodiments of the present disclosure can locate and segment the iris region in the iris image by performing target detection on the iris image, and obtain a segmentation result corresponding to the iris region. At the same time, it can also perform multi-scale feature extraction on the iris image. Fusion with the feature to obtain a high-precision feature map, and then use the segmentation result and the feature map to perform identity recognition of the iris image to determine whether each iris image corresponds to the same object.
  • the extracted low-level features and high-level features can be fully integrated by fusing features, so that the final iris features can take into account the texture features of the bottom layer and the classification features of the high-level, improving the accuracy of feature extraction, and can also use the segmentation results
  • the combination with the feature map only considers the characteristic part of the iris area, reduces the influence of other areas, and more accurately recognizes whether the iris image corresponds to the same object, and the detection result is higher.
  • Fig. 3 shows a flowchart of step S20 in an image processing method according to an embodiment of the present disclosure.
  • the detecting the position of the iris in the iris image and the segmentation result of the iris area in the iris image includes:
  • S21 Perform target detection processing on the iris image, and determine the iris position and the pupil position of the iris image;
  • S22 Perform the segmentation process on the iris image based on the determined iris position and the pupil position to obtain a segmentation result of the iris region in the iris image.
  • preprocessing may be performed on the iris image to obtain the iris feature of the iris image and the segmentation result of the iris region.
  • the target detection process of the iris image can be performed first through a neural network capable of performing target detection.
  • the neural network may be a convolutional neural network, which is trained to recognize the position of the iris and the position of the pupil in the iris image.
  • Fig. 4 shows a schematic diagram of preprocessing of an iris image according to an embodiment of the present disclosure.
  • A represents the iris image, and the iris position and pupil position in the iris image can be determined after the target detection process is executed.
  • B represents the image area corresponding to the iris position in the iris image.
  • the position of the iris and the position of the pupil obtained by performing target detection can be expressed as the position of the iris detection frame and the position of the pupil detection frame, and the position can be expressed as (x1, x2, y1, y2), where ( x1, y1) and (x2, y2) are the position coordinates of the two diagonal vertices of the detection frame of the iris or pupil, respectively, and the location of the area corresponding to the corresponding detection frame can be determined based on the coordinates of the two vertices.
  • the position of the detection frame may also be expressed in other forms, which is not specifically limited in the present disclosure.
  • the neural network that performs the target detection processing in the embodiment of the present disclosure may include: Faster R-CNN neural network (fast target recognition convolutional neural network) or Retina network (single-stage target detection network), but it is not as specific in the present disclosure. limited.
  • the segmentation of the iris area in the iris image can be further performed, so that the iris area can be segmented and distinguished from other parts such as eyelids and pupils.
  • the iris position and pupil position can be used to directly segment the iris area in the iris image, that is, the image area corresponding to the pupil position can be deleted from the image area corresponding to the iris position, and the image area of the iris position can be deleted.
  • the remaining image area is determined as the segmentation result of the iris area, and the iris area is assigned the mask value of the first identifier, and the remaining areas are assigned the mask value of the second identifier to obtain the mask map corresponding to the iris area.
  • This method has the characteristics of simplicity and convenience, and improves the processing speed.
  • the iris position, the pupil position, and the corresponding iris image can be input to a neural network for performing iris segmentation, and the neural network outputs a mask map corresponding to the iris region in the iris image.
  • the neural network for performing iris segmentation may be trained to be able to determine the iris region in the iris image and generate a corresponding mask map.
  • the neural network may also be a convolutional neural network, for example, PSPNet (Pyramid Scene Analysis Network) or Unet (U-shaped network), but it is not a specific limitation of the present disclosure.
  • C in Figure 4 represents a schematic diagram of the iris area corresponding to the iris image, where the black part represents the image area outside the iris area, the mask value of this part is the second identifier, the white part is the iris area, and the mask value of this part For the first logo.
  • the iris area can be accurately detected, and the accuracy of subsequent comparison processing can be improved.
  • the embodiment of the present disclosure can obtain the iris position in the iris image and the mask map of the iris region (segmentation). In the case of result), it is also possible to perform normalization processing on the image area and the mask image corresponding to the iris position, so that the normalized image area and the mask image are adjusted to the preset specifications.
  • the embodiment of the present disclosure can adjust the image area of the iris position and the mask map to a height of 64 pixels and a width of 512 pixels.
  • the specific dimensions of the aforementioned preset specifications are not specifically limited in this disclosure.
  • the characteristic processing of the corresponding image area can be performed based on the obtained iris position to obtain the iris characteristic.
  • the embodiment of the present disclosure can also perform multi-scale feature processing on the image region corresponding to the normalized iris position, so as to further improve the feature accuracy.
  • the following is an example of directly performing multi-scale feature extraction and multi-scale fusion processing on the image area corresponding to the iris position.
  • the multi-scale feature processing and multi-scale fusion processing of the image area corresponding to the normalized iris position are no longer To repeat the explanation, the process of the two is the same.
  • Fig. 5 shows a flowchart of step S30 in an image processing method according to an embodiment of the present disclosure.
  • the performing multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain the iris feature map corresponding to the iris image includes:
  • S31 Perform the multi-scale feature extraction process on the image area corresponding to the iris position in the iris image to obtain feature maps of multiple scales;
  • the feature extraction process may be performed first on the image region corresponding to the iris position in the iris image, where the feature extraction process may be performed using a feature extraction neural network, for example, a residual network or a pyramid feature may be used
  • the extraction network executes the feature extraction process to obtain feature maps of multiple scales corresponding to the image area where the iris position of the iris image is located.
  • Fig. 6 shows a schematic structural diagram of a neural network according to an image processing method implementing an embodiment of the present disclosure.
  • the network structure M represents the part of the neural network that performs feature extraction, which may be a residual network, such as ResNet18, but it is not a specific limitation of the present disclosure.
  • the iris image and the iris position of the iris image can be input to the feature extraction neural network, and the neural network is processed through the feature extraction to obtain the feature corresponding to the image area corresponding to the iris position of the iris image, that is, the feature of multiple scales.
  • the image area corresponding to the position of the iris may be firstly intercepted from the iris image, and the image area may be input to the feature extraction neural network to obtain feature maps of multiple scales, wherein the features of the multiple scales
  • the image can be output from different convolutional layers of the feature extraction neural network, and at least two feature maps of different scales can be obtained.
  • the obtained feature maps of multiple scales can include low-level feature information (the feature map obtained by using the convolutional layer in front of the network architecture) or high-level feature information (using the network architecture The feature map obtained by the subsequent convolutional layer), through the fusion of the above features, a more accurate and comprehensive iris feature can be obtained.
  • S32 Use the feature maps of the multiple scales to form at least one feature group, where the feature group includes feature maps of at least two scales in the feature maps of the multiple scales;
  • At least one feature group may be formed based on the feature maps of the multiple scales.
  • the feature maps of multiple scales can be regarded as a feature group, and subsequent feature fusion processing can be performed, or at least two feature groups can be formed, and each feature group can include at least two feature maps of different scales, and
  • the different feature groups formed by the embodiments of the present disclosure may include the same feature map, that is, any two feature groups may include at least one different feature map.
  • the multi-scale feature map obtained in step S31 may include F1, F2, and F3, and the scales of the three feature maps are different.
  • a first preset number of feature groups may be formed, and the first preset number may be an integer greater than or equal to 1, for example, the first preset number may take a value of 2 in the embodiment of the present disclosure.
  • each feature group can be assigned a second preset number of feature maps, where the second preset number of feature maps can be randomly selected from feature maps of multiple scales to form a feature group, and the selected feature map remains Can be grouped and selected by other features.
  • the second preset number may be an integer greater than or equal to 2.
  • the second preset number in the embodiment of the present disclosure may take a value of 2.
  • the feature maps in one feature group formed are F1 and F2, and the features in another feature group can be F1 and F3.
  • feature fusion processing may be performed on the feature maps in each feature group.
  • a spatial attention mechanism is adopted.
  • the convolution processing based on the attention mechanism can be realized through the spatial attention neural network, and the obtained feature map further highlights the important features.
  • the importance of each position of the spatial feature can be learned adaptively, and the attention coefficient of the feature object of each position can be formed. This coefficient can represent the interval of [0,1] The coefficient value.
  • the spatial attention neural network can be a network structure N.
  • grouping convolution and standard convolution processing can be further performed to further obtain the fusion feature of each feature map in each feature group, that is, the grouped feature map.
  • S34 Obtain an iris feature map corresponding to the iris image based on the grouped feature map corresponding to the feature grouping.
  • feature fusion can be performed on the grouped feature maps of different feature groups to obtain the iris features corresponding to the iris image.
  • the sum result of the grouped feature maps of each feature group can be used as the iris feature map, or the weighted sum of each grouped feature map can be used as the iris feature map, wherein the weighting coefficient of the weighted sum can be based on requirements And the setting of the scene, this disclosure does not specifically limit this.
  • the fusion can be performed for different feature maps.
  • the attention mechanism can further improve the attention of important features, and then the fusion processing of grouped feature maps based on different feature groups can further integrate the features of each part more comprehensively.
  • Fig. 7 shows a flowchart of step S33 in an image processing method according to an embodiment of the present disclosure.
  • the multi-scale feature fusion process is performed on the feature maps in the feature group based on the attention mechanism to obtain the The grouping feature map corresponding to the feature grouping, including:
  • S331 Perform a first convolution process on the connected feature maps of the feature maps of the at least two scales in the feature group to obtain a first sub-feature map;
  • connection processing on the feature maps in each feature group, such as concatenate in the channel direction, to obtain a connected feature map, as shown in FIG. 6, in which the convolution is enlarged.
  • SAFFM Neural Network of Attention Mechanism
  • the scale of the connection feature map obtained by the connection process can be expressed as (C, H, W), C represents the number of channels of the connection feature map, H represents the height of the connection feature map, and W represents the width of the connection feature map.
  • F1 and F2 in the feature maps in the above feature group can be connected, and the feature maps F1 and F3 in another feature group can be connected to obtain corresponding connection feature maps respectively.
  • the first convolution process can be performed on each connection feature map, such as using a 3*3 convolution kernel to perform the first convolution process, and then batch normalization and activation can also be performed Function processing to obtain the first sub-feature map corresponding to the connected feature map.
  • the scale of the first sub-feature map can be expressed as (C/2, H, W), and the parameters in the feature map can be reduced through the first convolution process , Which reduces the subsequent calculation cost.
  • S332 Perform second convolution processing and activation function processing on the first sub-feature map to obtain a second sub-feature map, where the second sub-feature map represents the attention coefficient corresponding to the first sub-feature map;
  • the second convolution process may be performed on the obtained first sub-feature map.
  • two convolution layers may be used to perform the second convolution process, one of which is convolution.
  • batch normalization and activation function processing are performed to obtain the first intermediate feature map.
  • the scale of the first intermediate feature map can be expressed as (C/8, H, W)
  • the convolution processing of the 1*1 convolution kernel is performed on the intermediate feature map through the second convolution layer to obtain the second intermediate feature map of (1, H, W).
  • dimensionality reduction processing can be performed on the first sub-feature map to obtain a single-channel second intermediate feature map.
  • the sigmoid function can be used to perform activation function processing on the second intermediate feature map. After the second intermediate feature map is processed by the sigmoid function, a second sub-feature map corresponding to the first sub-feature map can be obtained. Each element in the figure represents the attention coefficient of the feature value of each pixel in the first sub-feature map. The coefficient value can be a value in the range of [0,1].
  • product processing may be performed on the first sub-feature map and the second sub-feature map, such as multiplying corresponding elements. Then the product result is added to the first sub-characteristic map (add), that is, the corresponding elements are added to obtain the third sub-characteristic map.
  • the characteristic map output by SAFFM is the third characteristic map. Since the input feature groups are different, the third sub-feature map obtained is also different.
  • S334 Perform a third convolution process on the third sub-feature map to obtain a grouped feature map corresponding to the feature group.
  • a third convolution process may be performed on the third sub-feature map, and the third convolution process may include at least one of grouped convolution processing and standard convolution processing.
  • the third convolution process can further realize the further fusion of the feature information in each feature group.
  • the third convolution process can include grouped convolution (depthwise conv) and standard convolution with a 1*1 convolution kernel, where grouped convolution can speed up the convolution and at the same time improve the convolution features. Accuracy.
  • the grouped feature map corresponding to each feature group can be finally output.
  • the grouped feature map effectively integrates the feature information of each feature map in the feature group.
  • the weighted sum or addition of the grouped feature maps can be used to obtain the iris feature map corresponding to the iris image.
  • Fig. 8 shows a flowchart of step S40 in an image processing method according to an embodiment of the present disclosure.
  • the using the mask map and the iris feature map corresponding to the at least two iris images in the iris image group to perform the comparison processing includes:
  • S41 Determine a first position of the at least two iris images that is the same as the iris area by using the segmentation results respectively corresponding to the at least two iris images;
  • the segmentation result may be expressed as a mask image of the location of the iris region in the iris image. Based on this, it is possible to determine the first position of the iris region in the iris image to be compared according to the mask map of each iris image.
  • the first identifier in the mask image can indicate the position of the iris area. If the pixel points at the same position in the mask images of two iris images have the corresponding mask values as the first identifier, that is, It can indicate that the pixel point is located in the mask area in the two iris images, and based on the positions of all such pixel points, it can be determined that the two iris images are the first positions of the iris area.
  • the first position of the same iris region can be determined according to the product between the mask images of the two iris images, wherein the product result of the mask image is still the first identified pixel
  • the position of the point is the first position of the same iris area.
  • S42 Determine respectively the fourth sub-feature map corresponding to the first position in the iris feature map of the at least two iris images
  • the embodiment of the present disclosure can obtain the feature corresponding to the above-mentioned first position in the iris feature map of each iris image, that is, the fourth sub-feature map.
  • the embodiment of the present disclosure may determine the characteristic value of the corresponding pixel according to the coordinates of the first position, and form the fourth sub-characteristic map according to the determined characteristic value and the corresponding pixel.
  • the iris feature map of the two iris images to be compared can be obtained at the same position as the iris area. Performing the comparison of the two iris images according to this feature can reduce the characteristic information of the non-iris area. Influence and improve the accuracy of the comparison.
  • S43 Determine a comparison result of the at least two iris images according to the degree of association between the fourth sub-feature maps respectively corresponding to the at least two iris images.
  • the correlation degree between the fourth sub-feature maps corresponding to the two iris images to be compared can be obtained, and then the correlation degree can be obtained Determine the correlation between the two iris images to be compared, that is, the comparison result.
  • the above-mentioned correlation degree may be Euclidean distance, or may also be cosine similarity, which is not specifically limited in the present disclosure.
  • the comparison process of two iris images to be compared can be expressed as:
  • SD (f 1 , f 2 ) represents the comparison result (degree of association) between two iris images
  • m 1 and m 2 represent the mask images of the two iris images
  • f 1 and f 2 represent two iris images. Iris feature map of each iris image.
  • the two iris images to be compared it can be determined whether the two iris images correspond to the same person object according to the comparison result.
  • the correlation between the fourth sub-feature maps corresponding to the two iris images to be compared is greater than the first threshold, it can indicate that the correlation between the two iris images is high, and at this time, it can be determined that the The two iris images to be compared correspond to the same object.
  • the correlation between the fourth sub-feature maps corresponding to the two iris images to be compared is less than or equal to the first threshold, it may indicate that the correlation between the two iris images is revealed, and it is determined at this time
  • the two iris images to be compared correspond to different objects.
  • the first threshold may be a preset value, such as 70%, but it is not a specific limitation of the present disclosure.
  • the image processing method provided by the embodiment of the present disclosure can be implemented using a neural network, for example, can be implemented using the network structure shown in FIG. 6, and the process of training the neural network will be described below.
  • the training image group can be obtained.
  • the training images can include at least two iris images of human objects, each of which has at least one iris image, and the resolution, image quality, and size of each iris image can be different, so Can improve the applicability of neural networks.
  • the neural network can be used to perform image processing of the training images, and the grouped feature maps corresponding to the feature groups obtained by the image processing of each training image can be obtained. Then obtain the network loss of the neural network based on the obtained grouping feature map.
  • the network loss is less than the loss threshold, it can indicate that the detection accuracy of the neural network meets the requirements and can be applied.
  • the network loss is greater than or equal to the loss threshold , Can feedback and adjust the parameters of the neural network network, such as convolution parameters, etc., until the loss function obtained is less than the loss threshold.
  • the loss threshold may be a value set according to requirements, such as 0.1, but it is not a specific limitation of the present disclosure.
  • the embodiments of the present disclosure may determine the network loss according to the minimum degree of association between the iris feature maps of the same person object and the maximum degree of association between different person objects.
  • the loss function can be expressed as:
  • L total ⁇ 1 L 1 + ⁇ 1 L 2 .
  • Ls represents the network loss corresponding to the grouped feature maps obtained from the same feature fusion branch
  • P represents the total number of person objects
  • K represents the total number of iris images of each person object
  • s represents the number of feature grouping groups
  • m represents the common iris Area
  • B represents the number of columns in the grouping feature map
  • the MMSD function represents the degree of association between the features.
  • MMSD(f 1,s ,f 2,s ) represents the two training images
  • the grouped feature map of a training image f 1, s is the feature map obtained after column transposition
  • L total represents the weighted sum of the network loss corresponding to the grouped feature maps obtained for different feature fusion branches, that is, the network loss of the entire neural network
  • ⁇ 1 and ⁇ 1 respectively represent the weighting coefficient
  • L 1 and L 2 respectively represent the corresponding two groups Network loss.
  • the iris region in the iris image is located and segmented, and the segmentation result corresponding to the iris position and the iris region is obtained.
  • multi-scale feature extraction and extraction of the iris image can be performed. Multi-scale feature fusion to obtain a high-precision iris feature map, and then use the segmentation result and the iris feature map to perform identity recognition of the iris image to determine whether each iris image corresponds to the same object.
  • the extracted low-level features and high-level features can be fully integrated through multi-scale feature extraction and multi-scale fusion features, so that the finally obtained iris feature takes into account the texture features of the bottom layer and the classification features of the high-level, improving the accuracy of feature extraction
  • it can also use the combination of the segmentation result and the iris feature map to consider only the characteristic part of the iris area, reduce the influence of other areas, and more accurately identify whether the iris image corresponds to the same object, and the detection result is higher.
  • the embodiments of the present disclosure can adopt a spatial attention mechanism in a neural network to allow the network to adaptively learn iris features in view of the different importance of texture regions in the iris image.
  • the writing order of the steps does not mean a strict execution order but constitutes any limitation on the implementation process.
  • the specific execution order of each step should be based on its function and possibility.
  • the inner logic is determined.
  • the present disclosure also provides image processing devices, electronic equipment, computer-readable storage media, and programs, all of which can be used to implement any image processing method provided in the present disclosure.
  • image processing devices electronic equipment, computer-readable storage media, and programs, all of which can be used to implement any image processing method provided in the present disclosure.
  • Fig. 9 shows a block diagram of an image processing device according to an embodiment of the present disclosure. As shown in Fig. 9, the image processing device includes:
  • the acquiring module 10 is configured to acquire an iris image group, the iris image group including at least two iris images to be compared;
  • the detection module 20 is used to detect the position of the iris in the iris image and the segmentation result of the iris area in the iris image;
  • the feature processing module 30 is configured to perform multi-scale feature extraction and multi-scale feature fusion processing on the image region corresponding to the iris position to obtain an iris feature map corresponding to the iris image;
  • the comparison module 40 is configured to use the segmentation result and the iris feature map corresponding to the at least two iris images to perform a comparison process, and determine the to-be-compared based on the comparison result of the comparison process Whether the two iris images correspond to the same object.
  • the feature processing module is further configured to perform the multi-scale feature extraction process on the image region corresponding to the iris position in the iris image to obtain feature maps of multiple scales;
  • the feature group including feature maps of at least two scales in the feature maps of the multiple scales;
  • the iris feature map corresponding to the iris image is obtained based on the grouped feature map corresponding to the feature grouping.
  • the feature processing module is further configured to perform the multi-scale feature extraction process on the image region corresponding to the iris position in the iris image to obtain feature maps of multiple scales;
  • the feature group including feature maps of at least two scales in the feature maps of the multiple scales;
  • the iris feature map corresponding to the iris image is obtained based on the grouped feature map corresponding to the feature grouping.
  • the feature processing module is further configured to perform weighting and processing on the grouped feature map corresponding to each of the grouped features to obtain the iris feature map corresponding to the iris image.
  • the segmentation result includes a mask image corresponding to an iris region in the iris image, the first identifier in the mask image represents the iris region, and the first identifier in the mask image
  • the second mark indicates a location area outside the iris area.
  • the second detection module is further configured to perform target detection processing on the iris image, and determine the iris position and the pupil position of the iris image;
  • the segmentation process is performed on the iris image to obtain a segmentation result of the iris region in the iris image.
  • the detection module is further configured to perform normalization processing on the image area corresponding to the iris position of the iris image and the segmentation result respectively;
  • the feature processing module is also used for:
  • the comparison module is further configured to use the segmentation results respectively corresponding to the at least two iris images to determine the first position of the at least two iris images that are both iris regions;
  • the comparison result of the at least two iris images is determined according to the degree of association between the fourth sub-feature maps respectively corresponding to the at least two iris images.
  • the comparison module is further configured to determine that the at least two iris images respectively correspond to a fourth sub-feature map with a correlation degree greater than a first threshold.
  • the two iris images correspond to the same object.
  • the comparison module is further configured to determine that the degree of association between the fourth sub-feature maps corresponding to the at least two iris images is less than or equal to a first threshold. At least two iris images correspond to different objects.
  • the device includes a neural network
  • the neural network includes the acquisition module, the detection module, the feature processing module, and the comparison module.
  • the functions or modules contained in the device provided in the embodiments of the present disclosure can be used to execute the methods described in the above method embodiments.
  • the functions or modules contained in the device provided in the embodiments of the present disclosure can be used to execute the methods described in the above method embodiments.
  • the embodiments of the present disclosure also provide a computer-readable storage medium on which computer program instructions are stored, and the computer program instructions implement the above-mentioned method when executed by a processor.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium.
  • An embodiment of the present disclosure also provides an electronic device, including: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured as the above method.
  • the electronic device can be provided as a terminal, server or other form of device.
  • Fig. 10 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • the electronic device 800 may be a mobile phone, a computer, a digital broadcasting terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and other terminals.
  • the electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, and a sensor component 814 , And communication component 816.
  • the processing component 802 generally controls the overall operations of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
  • the processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the foregoing method.
  • the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components.
  • the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.
  • the memory 804 is configured to store various types of data to support operations in the electronic device 800. Examples of these data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc.
  • the memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable and Programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable and Programmable read only memory
  • PROM programmable read only memory
  • ROM read only memory
  • magnetic memory flash memory
  • flash memory magnetic disk or optical disk.
  • the power supply component 806 provides power for various components of the electronic device 800.
  • the power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the electronic device 800.
  • the multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes one or more touch sensors to sense touch, sliding, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure related to the touch or slide operation.
  • the multimedia component 808 includes a front camera and/or a rear camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
  • the audio component 810 is configured to output and/or input audio signals.
  • the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal.
  • the received audio signal may be further stored in the memory 804 or transmitted via the communication component 816.
  • the audio component 810 further includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, a button, and the like. These buttons may include, but are not limited to: home button, volume button, start button, and lock button.
  • the sensor component 814 includes one or more sensors for providing the electronic device 800 with various aspects of state evaluation.
  • the sensor component 814 can detect the on/off status of the electronic device 800 and the relative positioning of the components.
  • the component is the display and the keypad of the electronic device 800.
  • the sensor component 814 can also detect the electronic device 800 or the electronic device 800.
  • the position of the component changes, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration/deceleration of the electronic device 800, and the temperature change of the electronic device 800.
  • the sensor component 814 may include a proximity sensor configured to detect the presence of nearby objects when there is no physical contact.
  • the sensor component 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • the communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices.
  • the electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof.
  • the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 further includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field-available A programmable gate array (FPGA), controller, microcontroller, microprocessor, or other electronic components are implemented to implement the above methods.
  • ASIC application-specific integrated circuits
  • DSP digital signal processors
  • DSPD digital signal processing devices
  • PLD programmable logic devices
  • FPGA field-available A programmable gate array
  • controller microcontroller, microprocessor, or other electronic components are implemented to implement the above methods.
  • a non-volatile computer-readable storage medium such as the memory 804 including computer program instructions, which can be executed by the processor 820 of the electronic device 800 to complete the foregoing method.
  • the embodiments of the present disclosure also provide a computer program product, including computer readable code, and when the computer readable code runs on the device, the processor in the device executes instructions for implementing the method provided in any of the above embodiments.
  • the computer program product can be specifically implemented by hardware, software, or a combination thereof.
  • the computer program product is specifically embodied as a computer storage medium.
  • the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc. Wait.
  • SDK software development kit
  • Fig. 11 shows a block diagram of another electronic device according to an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server. 11, the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932, for storing instructions executable by the processing component 1922, such as application programs.
  • the application program stored in the memory 1932 may include one or more modules each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above-described methods.
  • the electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to the network, and an input output (I/O) interface 1958 .
  • the electronic device 1900 can operate based on an operating system stored in the memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or the like.
  • a non-volatile computer-readable storage medium is also provided, such as the memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to complete the foregoing method.
  • the present disclosure may be a system, method and/or computer program product.
  • the computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure.
  • the computer-readable storage medium may be a tangible device that can hold and store instructions used by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Non-exhaustive list of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) Or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a printer with instructions stored thereon
  • RAM random access memory
  • ROM read-only memory
  • EPROM erasable programmable read-only memory
  • flash memory flash memory
  • SRAM static random access memory
  • CD-ROM compact disk read-only memory
  • DVD digital versatile disk
  • memory stick floppy disk
  • mechanical encoding device such as a printer with instructions stored thereon
  • the computer-readable storage medium used here is not interpreted as the instantaneous signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (for example, light pulses through fiber optic cables), or through wires Transmission of electrical signals.
  • the computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to various computing/processing devices, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • the network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network, and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing/processing device .
  • the computer program instructions used to perform the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or in one or more programming languages.
  • Source code or object code written in any combination, the programming language includes object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as "C" language or similar programming languages.
  • Computer-readable program instructions can be executed entirely on the user's computer, partly on the user's computer, executed as a stand-alone software package, partly on the user's computer and partly executed on a remote computer, or entirely on the remote computer or server carried out.
  • the remote computer can be connected to the user's computer through any kind of network-including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect to the user's computer) connection).
  • LAN local area network
  • WAN wide area network
  • an electronic circuit such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), can be customized by using the status information of the computer-readable program instructions.
  • FPGA field programmable gate array
  • PDA programmable logic array
  • the computer-readable program instructions are executed to realize various aspects of the present disclosure.
  • These computer-readable program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine that makes these instructions when executed by the processor of the computer or other programmable data processing device , A device that implements the functions/actions specified in one or more blocks in the flowcharts and/or block diagrams is produced. It is also possible to store these computer-readable program instructions in a computer-readable storage medium. These instructions make computers, programmable data processing apparatuses, and/or other devices work in a specific manner. Thus, the computer-readable medium storing the instructions includes An article of manufacture, which includes instructions for implementing various aspects of the functions/actions specified in one or more blocks in the flowchart and/or block diagram.
  • each block in the flowchart or block diagram may represent a module, program segment, or part of an instruction, and the module, program segment, or part of an instruction contains one or more components for realizing the specified logical function.
  • Executable instructions may also occur in a different order than the order marked in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, or they can sometimes be executed in the reverse order, depending on the functions involved.
  • each block in the block diagram and/or flowchart, and the combination of the blocks in the block diagram and/or flowchart can be implemented by a dedicated hardware-based system that performs the specified functions or actions Or it can be realized by a combination of dedicated hardware and computer instructions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Human Computer Interaction (AREA)
  • Ophthalmology & Optometry (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)
  • Collating Specific Patterns (AREA)

Abstract

本公开涉及一种图像处理方法及装置、电子设备和存储介质，所述方法包括获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述至少两个虹膜图像是否对应于同一对象。本公开实施例可实现虹膜图像的精确比对。

Description

图像处理方法及装置、电子设备和存储介质
本公开要求在2019年9月26日提交中国专利局、申请号为201910919121.9、申请名称为“图像处理方法及装置、电子设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及计算机视觉技术领域，尤其涉及一种图像处理方法及装置、电子设备和存储介质。
背景技术
虹膜识别技术是利用虹膜终身不变的稳定性和独一无二的差异性的特点来进行身份认证的。虹膜识别的优越性使得其在金融、电子商务、安全保卫、出入境控制等各个方面都具有极大的应用前景。
目前的虹膜识别算法一般都是利用滤波器来对虹膜进行特征提取。
发明内容
本公开提出了一种图像处理技术方案。
根据本公开的一方面，提供了一种图像处理方法，其包括：获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；利用所述虹膜图像组中的所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述至少两个虹膜图像是否对应于同一对象。基于上述配置，可以利用多尺度特征提取提取多个尺度的特征信息，例如可以同时得到底层和高层的特征信息，而后通过多尺度特征融合，得到的特征图精确度更高，进而实现更准确的比对，提高比对结果的准确性。
在一些可能的实施方式中，所述对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图，包括：对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。基于上述配置，可以实现得到的多个尺度的特征图的分组，并进一步引入注意力机制，确定相应分组的分组特征图，进一步提高得到的虹膜特征图的精度。
在一些可能的实施方式中，所述基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图，包括：对所述特征分组中的所述至少两个尺度的特征图的连接特征图执行第一卷积处理，得到第一子特征图；对所述第一子特征图执行第二卷积处理以及激活函数处理，得到第二子特征图，所述第二子特征图表示所述第一子特征图对应的注意力系数；将所述第一子特征图和所述第二子特征图的乘积结果与所述第一子特征图相加，得到第三子特征图；对所述第三子特征图执行第三卷积处理，得到所述特征分组对应的分组特征图。
在一些可能的实施方式中，所述基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图，包括：对每个所述分组特征对应的所述分组特征图执行加权和处理，得到所述虹膜图像对应的虹膜特征图。通过加权和的方式融合各分组的分组特征，实现特征信息的有效融合。
在一些可能的实施方式中，所述分割结果包括所述虹膜图像中虹膜区域对应的掩码图，所述掩码图中的第一标识表示所述虹膜区域，所述掩码图中的第二标识表示所述虹膜区域以外的位置区域。基于上述配置，通过掩码图表示分割结果，更加直观且处理方便。
在一些可能的实施方式中，所述检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果，包括：对所述虹膜图像执行目标检测处理，确定所述虹膜图像的虹膜位置以及瞳孔位置；基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图像中虹膜区域的分割结果。基于上述配置，可以准确的确定出虹膜图像中虹膜所对应的检测位置，以及虹膜区域的分割结果。
在一些可能的实施方式中，所述检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果，还包括：分别对所述虹膜图像的虹膜位置对应的图像区域和所述分割结果执行归一化处理；所述对所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图，还包括：对归一化处理后的所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图。基于上述配置，可以在对虹膜位置的图像区域以及分割结果执行归一化处理，可以提高适用性。
在一些可能的实施方式中，利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，包括：利用所述至少两个虹膜图像分别对应的分割结果，确定所述至少虹膜图像中同为虹膜区域的第一位置；分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；根据所述至少两个虹膜图像分别对应的第四子特征图之间的关联度，确定所述至少两个虹膜图像的比对结果。本公开实施例可以利用不同虹膜图像分别对应的分割结果，确定比对的虹膜图像中同为虹膜区域的位置，并利用该位置对应的特征进行比对，得到比对结果。降低了虹膜区域以外的区域特征的干扰，提高比对精度。
在一些可能的实施方式中，所述基于比对结果确定所述至少两个虹膜图像是否对应于同一对象，包括：在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，确定所述至少两个虹膜图像对应于同一对象。基于上述配置，通过第一阈值的设置，可以灵活的适应不同的场景，并能够方便的获得比对结果。
在一些可能的实施方式中，所述基于比对结果确定所述至少两个虹膜图像是否对应于同一对象，还包括：在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，确定所述至少两个虹膜图像对应于不同对象。
在一些可能的实施方式中，所述图像处理方法通过卷积神经网络实现。基于上述配置，可以通过神经网络精确且方便快速的得到两个虹膜图像的比对结果。
根据本公开的第二方面，提供了一种图像处理装置，其包括：获取模块，用于获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；检测模块，用于检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；特征处理模块，用于对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；比对模块，用于利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述至少两个虹膜图像是否对应于同一对象。
在一些可能的实施方式中，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述特征处理模块还用于对每个所述分组特征对应的所述分组特征图执行加权和处理，得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述分割结果包括所述虹膜图像中虹膜区域对应的掩码图，所述掩码图中的第一标识表示所述虹膜区域，所述掩码图中的第二标识表示所述虹膜区域以外的位置区域。
在一些可能的实施方式中，所述检测二模块还用于对所述虹膜图像执行目标检测处理，确定所述虹膜图像的所述虹膜位置以及瞳孔位置；
基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图 像中虹膜区域的分割结果。
在一些可能的实施方式中，所述检测模块还用于分别对所述虹膜图像的虹膜位置对应的图像区域和所述分割结果执行归一化处理；
所述特征处理模块还用于：对归一化处理后的所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述比对模块还用于利用所述至少两个虹膜图像分别对应的分割结果，确定所述至少两个虹膜图像中同为虹膜区域的第一位置；
分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；
根据所述至少两个虹膜图像分别对应的第四子特征图之间的关联度，确定所述至少两个虹膜图像的比对结果。
在一些可能的实施方式中，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，确定所述至少两个虹膜图像对应于同一对象。
在一些可能的实施方式中，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，确定所述至少两个虹膜图像对应于不同对象。
在一些可能的实施方式中，所述装置包括神经网络，所述神经网络包括所述获取模块、所述检测模块、所述特征处理模块以及所述比对模块。
根据本公开的第三方面，提供了一种电子设备，其包括：
处理器；
用于存储处理器可执行指令的存储器；
其中，所述处理器被配置为调用所述存储器存储的指令，以执行第一方面中任意一项所述的方法。
根据本公开的第四方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现第一方面中任意一项所述的方法。
根据本公开的第五方面，提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现所述图像处理方法。
在本公开实施例中，通过对虹膜图像执行预处理，对虹膜图像中的虹膜区域进行定位和分割，得到虹膜位置和虹膜的分割结果，同时还可以对虹膜图像执行多尺度特征提取和多尺度特征融合，获得高精度的虹膜特征图，而后利用分割结果与虹膜特征图执行虹膜图像的身份识别，确定各虹膜图像是否对应于同一对象。通过上述配置，可以通过多尺度特征提取和多尺度特征融合的方式使得提取的底层特征以及高层特征充分的融合，使得最后得到的虹膜特征兼顾底层的纹理特征和高层的分类特征，提高特征提取精度，同时还可以利用分割结果和虹膜特征图的结合，仅考虑虹膜区域的特征部分，减少其他区域的影响，更加精确的识别出虹膜图像是否对应于同一对象，检测结果较高。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。
根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1示出根据本公开实施例的一种图像处理方法的流程图；
图2示出根据本公开实施例的一种图像处理方法的过程示意图；
图3示出根据本公开实施例的一种图像处理方法中步骤S20的流程图；
图4示出根据本公开实施例的虹膜图像的预处理的示意图；
图5示出根据本公开实施例的一种图像处理方法中步骤S30的流程图；
图6示出根据实现本公开实施例的图像处理方法的神经网络的结构示意图；
图7示出根据本公开实施例的一种图像处理方法中步骤S33的流程图；
图8示出根据本公开实施例的一种图像处理方法中步骤S40的流程图；
图9示出根据本公开实施例的一种图像处理装置的框图；
图10示出根据本公开实施例的一种电子设备的框图；
图11示出根据本公开实施例的另一种电子设备的框图。
具体实施方式
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。
本公开实施例提供了一种图像处理方法，该图像处理方法可以用于通过虹膜图像对应的虹膜特征，区分识别虹膜图像相应的对象是否为相同对象，如是否为同一个人物对象的虹膜图像。图像处理方法的执行主体可以是图像处理装置，例如，图像处理方法可以由终端设备或服务器或其它处理设备执行，其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等，服务器可以为本地服务器也可以为云端服务器。在一些可能的实现方式中，该图像处理方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。
图1示出根据本公开实施例的一种图像处理方法的流程图，如图1所示，所述图像处理方法包括：
S10：获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；
在一些可能的实施方式中，在计算机视觉领域中，可以通过虹膜图像执行身份验证，用以识别虹膜图像对应的对象的身份，或者确定相应的对象是否具有权限。本公开实施例可以通过对虹膜图像进行特征处理，并基于得到的特征实现虹膜图像的比对，确认虹膜图像对应的对象是否为同一个对象。在其他实施例中，还可以进一步根据确定的虹膜图像是否对应于同一对象，执行相应的验证操作。
本公开实施例可以首先获得需要执行待比对的虹膜图像，该待比对的虹膜图像构成虹膜图像组，可以获取至少两个虹膜图像。例如，本公开实施例的待比对的虹膜图像可以通过虹膜相机采集得到，或者也可以通过其他设备传输接收得到，或者也可以从存储器中读取得到，上述仅为示例性说明，本公开对此不作具体限定。
S20：检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；
在一些可能的实施方式中，可以首先对虹膜图像执行预处理，其中预处理可以包括对虹膜图像中的虹膜和瞳孔进行定位，确定虹膜和瞳孔的位置。其中，虹膜位置和瞳孔位置可以分别表示为虹膜的检测框以及瞳孔的检测框对应的位置。基于检测到的虹膜图像中虹膜和瞳孔的位置，可以进一步对虹膜区域执行分割处理，得到对应的分割结果，其中分割结果可以表示成掩码图。
其中，掩码图可以表示为向量或者矩阵形式，掩码图可以和虹膜图像的像素点一一对应。掩码图中可以包括第一标识和第二标识，其中第一标识表示对应的像素点为虹膜区域，第二标识表示对应的像素点为非虹膜区域，例如第一标识可以为“1”，第二标识可以为“0”，从而可以通过掩码图中第一标识的像素点的位置构成的区域确定虹膜所在的区域。
S30：对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；
在一些可能的实施方式中，在确定虹膜图像中的虹膜位置的情况下，可以对该虹膜位置对应的图像区域执行多尺度特征提取处理，例如，可以得到至少两个尺度的特征图，而后通过对特征图执行卷积处理，可以实现特征的融合，进而得到虹膜图像的虹膜特征图。
在一些可能的实施方式中，特征提取处理的过程中可以得到虹膜位置对应的图像区域的多个不同尺度的特征图，例如，可以通过残差网络执行特征提取处理，而后对该多个尺度的特征图执行至少一次的卷积处理，得到融合了不同尺度的特征的虹膜特征图。其中，通过多尺度特征提取可以同时获得底层和高层的特征信息，通过多尺度的特征融合可以对底层和高层特征信息有效的融合，提高虹膜特征图的精确度。
在一些可能的实施方式中，也可以利用注意力机制，针对不同的特征得到不同的注意力系数，注意力系数可以表示特征的重要度，利用该注意力系数执行特征融合可以得到更健壮和更具区分力的特征。
S40：利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述至少两个虹膜图像是否对应于同一对象。
在一些可能的实施方式中，可以根据需要比对的至少两个虹膜图像的分割结果(如掩码图)，得到两个虹膜图像中均为虹膜区域的位置，基于虹膜区域的位置对应的特征之间的距离，可以得到该至少两个虹膜图像的比对结果。其中，如果该距离小于第一阈值，则表明比对的两个虹膜图像对应于同一对象，即属于同一对象的虹膜图像。否则，如果该距离大于或者等于第一阈值，则表明该两个虹膜图像不属于同一对象。在此需要说明的是，当虹膜图像组中包括三个或三个以上待比对的虹膜图像的情况下，可以分别比对任意两个虹膜图像，确定该任意两个虹膜图像是否对应于同一对象，并根据每个虹膜图像的比对结果确定虹膜图像组中同为一个对象的虹膜图像，同时还可以统计虹膜图像组中的虹膜图像对应的对象的数量。
下述以两个虹膜图像为例进行说明，执行多个虹膜图像的比对的原理与下述相同，不作重复说明。图2示出根据本公开实施例的一种图像处理方法的过程示意图，其中，可以首先获取两个虹膜图像A和B，对该两个虹膜图像执行预处理，可以得到虹膜图像中的虹膜位置、瞳孔位置以及虹膜区域对应的分割结果，如掩码图，而后可以利用虹膜特征提取模块执行虹膜位置对应的图像区域的特征提取和融合处理，得到虹膜图像的虹膜特征图，进一步利用比对模块基于掩码图和相应的虹膜特征图，得到图像A和图像B是否为同一个对象的比对结果(score)。
基于上述配置，本公开实施例可以通过对虹膜图像执行目标检测，对虹膜图像中的虹膜区域进行定位和分割，得到与虹膜区域对应的分割结果，同时还可以对虹膜图像执行多尺度的特征提取和特征融合，获得高精度的特征图，而后利用分割结果与特征图执行虹膜图像的身份识别，确定各虹膜图像是否对应于同一对象。通过上述配置，可以通过融合特征的方式使得提取的底层特征以及高层特征充分的融合，使得最后得到的虹膜特征兼顾底层的纹理特征和高层的分类特征，提高特征提取精度，同时还可以利用分割结果和特征图的结合，仅考虑虹膜区域的特征部分，减少其他区域的影响，更加精确的识别出虹膜图像是否对应于同一对象，检测结果较高。
下面结合附图对本公开实施例进行详细说明。本公开实施例在得到待比对的虹膜图像的情况下，可以虹膜图像执行目标检测处理，得到虹膜区域和瞳孔区域对应的位置，并进一步得到虹膜区域对应的分割掩码图。图3示出根据本公开实施例的一种图像处理方法中步骤S20的流程图。其中，所述检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果，包括：
S21：对所述虹膜图像执行目标检测处理，确定所述虹膜图像的所述虹膜位置以及瞳孔位置；
S22：基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图像中虹膜区域的分割结果。
在一些可能的实施方式中，在得到虹膜图像的情况下，可以对虹膜图像执行预处理，得到上述虹膜图像的虹膜特征，以及虹膜区域的分割结果。其中，可以首先通过能够执行目标检测的神经网络执行虹膜图像的目标检测处理。其中该神经网络可以为卷积神经网络，其为经过训练能够识别出虹膜图像中的虹膜位置以及瞳孔位置。图4示出根据本公开实施例的虹膜图像的预处理的示意图。其中，A 表示虹膜图像，执行目标检测处理后可以确定虹膜图像中的虹膜位置以及瞳孔位置。B表示虹膜图像中的虹膜位置对应的图像区域，对于瞳孔位置对应的区域图中并未标出，在实际应用和检测过程中，同时可以检测出瞳孔位置对应的区域。如上述实施例所述，执行目标检测得到的虹膜的位置以及瞳孔的位置可以表示为虹膜的检测框以及瞳孔检测框的位置，该位置可以表示为(x1，x2，y1，y2)，其中(x1，y1)和(x2，y2)分别虹膜或者瞳孔的检测框的两个对角顶点的位置坐标，基于该两个顶点坐标可以确定相应的检测框对应的区域位置。在其他实施例中，检测框的位置也可以表示成其他形式，本公开对此不作具体限定。
另外，本公开实施例中执行目标检测处理的神经网络可以包括：Faster R-CNN神经网络(快速目标识别卷积神经网络)或者Retina网络(单级目标检测网络)，但不作为本公开的具体限定。
在得到虹膜图像中的虹膜的位置以及瞳孔的位置的情况下，可以进一步执行虹膜图像中虹膜区域的分割，从而可以将虹膜区域与眼皮、瞳孔等其他部位分割区分。
在一个示例中，可以利用虹膜位置和瞳孔位置，直接对虹膜图像中的虹膜区域进行分割处理，即，可以从虹膜位置对应的图像区域中删除瞳孔位置对应的图像区域，将虹膜位置的图像区域中余下的图像区域确定为虹膜区域的分割结果，并为该虹膜区域赋予第一标识的掩码值，其余区域赋予第二标识的掩码值，得到虹膜区域对应的掩码图。该方式具有简单方便的特点，提高处理速度。
在另一个示例中，可以将虹膜位置、瞳孔位置以及对应的虹膜图像输入至用于执行虹膜分割的神经网络中，通过该神经网络输出虹膜图像中虹膜区域对应的掩码图。其中，该执行虹膜分割的神经网络可以为经过训练能够实现对于虹膜图像中的虹膜区域的确定，并生成对应的掩码图。该神经网络同样可以为卷积神经网络，例如可以为PSPNet(金字塔场景分析网络)或者Unet(U型网络)，但不作为本公开的具体限定。图4中的C表示与虹膜图像对应的虹膜区域的示意图，其中黑色部分表示虹膜区域以外的图像区域，该部分的掩码值为第二标识，白色部分为虹膜区域，该部分的掩码值为第一标识。通过该方式可以精确的检测出虹膜区域，提高后续比对处理的精确度。
在一些可能的实施方式中，由于得到的虹膜图像以及检测到的虹膜区域的掩码图的大小可能不同，本公开实施例可以在得到虹膜图像中的虹膜位置以及虹膜区域的掩码图(分割结果)的情况下，还可以对虹膜位置对应的图像区域以及掩码图执行归一化处理，使得归一化后的图像区域和掩码图调整为预设规格。例如，本公开实施例可以将虹膜位置的图像区域以及掩码图调整为高度64像素，宽度512像素。但上述预设规格的具体尺寸本公开不作具体限定。
在得到虹膜图像中虹膜的位置以及对应的分割结果的情况下，可以基于得到的虹膜位置执行相应图像区域的特征处理，得到虹膜特征。另外，本公开实施例还可以对归一化后的虹膜位置对应的图像区域执行多尺度的特征处理，进一步提高特征精度。下述以直接对虹膜位置对应的图像区域执行多尺度特征提取和多尺度融合处理为例进行说明，对于归一化后的虹膜位置对应的图像区域的多尺度特征处理和多尺度融合处理不再做重复说明，二者的处理过程相同。
图5示出根据本公开实施例的一种图像处理方法中步骤S30的流程图。其中所述对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图，包括：
S31：对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
在一些可能的实施方式中，可以首先对虹膜图像中所述虹膜位置对应的图像区域执行特征提取处理，其中，可以利用特征提取神经网络执行该特征提取处理，例如可以利用残差网络或者金字塔特征提取网络执行该特征提取处理，得到虹膜图像的虹膜位置所在的图像区域对应的多个尺度的特征图。图6示出根据实现本公开实施例的图像处理方法的神经网络的结构示意图。其中，网络结构M表示执行特征提取的神经网络部分，其可以为残差网络，如ResNet18，但不作为本公开的具体限定。
其中，在一个示例中，可以将虹膜图像以及虹膜图像的虹膜位置输入至特征提取神经网络，通过特征提取处理神经网络得到虹膜图像的虹膜位置对应的图像区域对应的特征，即多个尺度的特征图。或者，在另一个示例中，可以首先从虹膜图像中截取与虹膜位置对应的图像区域，将该图像区域输入 至特征提取神经网络，得到多个尺度的特征图，其中，该多个尺度的特征图可以分别通过特征提取神经网络的不同卷积层输出，可以得到至少两个不同尺度的特征图，如在一个示例中可以分别得到三个尺度的特征图，该三个特征图的尺度各不相同。另外，为了得到全面的特征信息，得到的多个尺度的特征图中可以包括底层的特征信息(利用网络架构前面的卷积层得到的特征图)，也可以包括高层的特征信息(利用网络架构后面的卷积层得到的特征图)，通过上述特征的融合，可以得到更精确且全面的虹膜特征。
S32：利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
在一些可能的实施方式中，在得到多个尺度的特征图的情况下，可以基于该多个尺度的特征图形成至少一个特征分组。
其中，可以将该多个尺度的特征图作为一个特征分组，执行后续的特征融合处理，或者也可以形成至少两个特征分组，每个特征分组中可以包括至少两个不同尺度的特征图，而且，本公开实施例形成的不同特征分组中可以包括相同的特征图，即任意两个特征分组中可以至少包括一个不同的特征图。例如步骤S31得到的多尺度特征图可以包括F1、F2和F3，该三个特征图的尺度各不相同。在形成特征分组时，可以形成第一预设数量个特征分组，该第一预设数量可以为大于或者等于1的整数，如本公开实施例中第一预设数量可以取值为2。继而可以为每个特征分组分配第二预设数量个特征图，其中可以随机的从多个尺度的特征图中选择出第二预设数量个特征图组成一个特征分组，被选择的特征图依然可以被其他特征分组选择。其中第二预设数量可以大于或者等于2的整数，如本公开实施例中的第二预设数量可以取值为2。例如形成的一个特征分组中的特征图为F1和F2，另一个特征分组中的特征可以为F1和F3。
S33：基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
在一些可能的实施方式中，在得到多个尺度的特征图的特征分组的情况下，可以对每个特征分组内的特征图执行特征融合处理。在融合的过程中，考虑到虹膜特征在不同位置上的重要性不同，采用了空间注意力机制。其中，可以通过空间注意力神经网络实现基于注意力机制的卷积处理，得到的特征图中进一步突出了重要的特征。在该空间注意力神经网络的训练过程中可以自适应的学到空间特征每个位置的重要性，形成与每个位置的特征对象的注意力系数，该系数可以表示[0,1]区间的系数值。如图6所示，空间注意力神经网络可以为网络结构N。
在利用注意力机制的神经网络执行卷积处理后，还可以进一步执行分组卷积、标准卷积处理，进一步得到每个特征分组中的各特征图的融合特征，即分组特征图。
S34：基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，在得到每个特征分组中的特征图的融合特征(分组特征图)的情况下，可以对不同特征分组的分组特征图执行特征融合，得到虹膜图像对应的虹膜特征图。例如，本公开实施例可以将每个特征分组的分组特征图的加和结果作为虹膜特征图，或者也可以将各分组特征图的加权和作为虹膜特征图，其中加权和的加权系数可以根据需求和场景设定，本公开对此不作具体限定。
通过上述方式可以分别针对不同的特征图执行融合，其中基于注意力机制可以进一步提高重要特征的关注度，而后基于不同特征分组的分组特征图的融合处理，进一步更全面的融合各部分的特征。
下面对上述特征融合的过程进行详细说明。图7示出根据本公开实施例的一种图像处理方法中步骤S33的流程图，所述基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图，包括：
S331：对所述特征分组中的所述至少两个尺度的特征图的连接特征图执行第一卷积处理，得到第一子特征图；
在一些可能的实施方式中，可以首先对每个特征分组中的特征图执行连接处理，如在通道方向上进行连接(concatenate)，得到连接特征图，如图6所示，其中放大了卷积注意力机制的神经网络(SAFFM)。通过连接处理得到的连接特征图的尺度可以表示为(C，H，W)，C表示连接特征图的 通道数，H表示连接特征图的高度，W表示连接特征图的宽度。例如，可以对上述特征分组中的特征图中的F1和F2进行连接，以及对另一个特征分组中的特征图F1和F3进行连接，分别得到对应的连接特征图。
在得到连接特征图的情况下，可以分别对各连接特征图执行第一卷积处理，如利用3*3的卷积核执行该第一卷积处理，而后还可以执行批归一化以及激活函数处理，得到与连接特征图对应的第一子特征图，该第一子特征图的尺度可以表示为(C/2，H，W)，通过第一卷积处理可以减少特征图中的参数，减少了后续的计算成本。
S332：对所述第一子特征图执行第二卷积处理以及激活函数处理，得到第二子特征图，所述第二子特征图表示所述第一子特征图对应的注意力系数；
在一些可能的实施方式中，可以对得到的第一子特征图执行第二卷积处理，如图6所示，可以分别采用两个卷积层执行该第二卷积处理，其中一个卷积层通过1*1的卷积核处理后，执行批归一化以及激活函数处理，得到第一中间特征图，该第一中间特征图的尺度可以表示为(C/8，H，W)，而后通过第二个卷积层对该中间特征图执行1*1卷积核的卷积处理，得到(1，H，W)的第二中间特征图。通过上述第二卷积处理，可以对第一子特征图执行降维处理，得到单通道的第二中间特征图。
进一步可以对该第二中间特征图使用sigmoid函数执行激活函数处理，通过sigmoid函数对第二中间特征图处理后，可以得到与第一子特征图对应的第二子特征图，该第二子特征图中的各元素表示第一子特征图中每个像素点的特征值的注意力系数。该系数值可以为[0,1]范围内的数值。
S333：将所述第一子特征图和第二子特征图的乘积结果与所述第一子特征图相加，得到第三子特征图；
在一些可能的实施方式中，在得到表示注意力系数的第二子特征图的情况下，可以对第一子特征图和第二子特征图执行乘积处理(mul)，如对应元素相乘。而后将该乘积结果与第一子特征图相加(add)，即对应元素相加，得到第三子特征图，如图6所示，通过SAFFM输出的特征图即为第三特征图。由于输入的特征分组不同，得到的第三子特征图也不相同。
S334：对所述第三子特征图执行第三卷积处理，得到所述特征分组对应的分组特征图。
在得到第三子特征图的情况下，可以对该第三子特征图执行第三卷积处理，该第三卷积处理可以包括分组卷积处理、标准卷积处理中的至少一种。通过第三卷积处理可以进一步实现每个特征分组中的特征信息的进一步融合。如图6所示，第三卷积处理可以包括分组卷积(depth wise conv)和1*1的卷积核的标准卷积，其中分组卷积可以加快卷积速度，同时提高卷积特征的精度。通过第三卷积处理可以最终输出每个特征分组对应的分组特征图。该分组特征图有效的融合了特征分组中各特征图的特征信息。在得到每个特征分组的分组特征图的情况下，可以利用各分组特征图的加权和或者加和(add)得到对应虹膜图像的虹膜特征图。
在得到虹膜图像的虹膜特征图的情况下，可以进一步结合分割结果执行虹膜图像之间的比对处理。图8示出根据本公开实施例的一种图像处理方法中步骤S40的流程图。所述利用所述虹膜图像组中的所述至少两个虹膜图像分别对应的掩码图和虹膜特征图，执行比对处理，包括：
S41：利用所述至少两个虹膜图像分别对应的分割结果，确定所述至少两个虹膜图像中同为虹膜区域的第一位置；
在一些可能的实施方式中，分割结果可以表示为虹膜图像中虹膜区域所在位置的掩码图。基于此可以根据每个虹膜图像的掩码图，确定待比对的虹膜图像中均为虹膜区域的第一位置。如上述实施例所述，掩码图中的第一标识可以表示虹膜区域所在的位置，如果两个虹膜图像的掩码图中相同位置的像素点对应的掩码值均为第一标识，即可以表示该像素点在两个虹膜图像中均位于掩模区域内，基于所有这样的像素点的位置即可以确定两个虹膜图像均为虹膜区域的第一位置。
或者，在其他实施例中，也可以根据两个虹膜图像的掩码图之间的乘积，确定同为虹膜区域的第一位置，其中，掩码图的乘积结果中仍然为第一标识的像素点的位置，即表示同为虹膜区域的第一位置。
S42：分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；
在确定上述第一位置的情况下，本公开实施例可以得到每个虹膜图像的虹膜特征图中上述第一位置对应的特征，即第四子特征图。本公开实施例可以根据第一位置的坐标确定对应像素点的特征值，根据确定的特征值以及对应的像素点，形成第四子特征图。或者也可以利用上述两个虹膜图像的掩码图的乘积结果与每个虹膜特征图的乘积，得到该虹膜特征图对应的第四子特征图。
通过上述配置，可以得到两个待比对的虹膜图像的虹膜特征图中同为虹膜区域的位置处的特征，根据该特征执行两个虹膜图像的比对，可以减少非虹膜区域的特征信息的影响，提高比对精度。
S43：根据所述至少两个虹膜图像分别对应的第四子特征图之间的关联度，确定所述至少两个虹膜图像的比对结果。
在得到同为虹膜区域的虹膜特征(第四子特征图)的情况下，可以得到获得待比对的两个虹膜图像对应的第四子特征图之间的关联度，继而可以将该关联度确定为待比对的两个虹膜图像之间的关联度，即比对结果。本公开实施例中，上述关联度可以为欧式距离，或者也可以为余弦相似度，本公开对此不作具体限定。
在一个示例中，在关联度为欧式距离的情况下，两个待比对的虹膜图像的比对过程可以表示为：
Figure PCTCN2019121695-appb-000001
其中，SD(f 1,f 2)表示两个虹膜图像之间的比对结果(关联度)，m 1和m 2分别表示两个虹膜图像的掩码图，f 1和f 2分别表示两个虹膜图像的虹膜特征图。
在得到上述两个待比对的虹膜图像的比对结果的情况下，可以根据比对结果确定两个虹膜图像是否对应于同一人物对象。其中，在待比对的两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，可以表示两个虹膜图像的关联度较高，此时可以确定所述待比对的两个虹膜图像对应于同一对象。否则，在所述待比对的两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，可以表示两个虹膜图像的关联度交底，此时确定所述待比对的两个虹膜图像对应于不同对象。第一阈值可以为预先设定的值，如可以为70％，但不作为本公开的具体限定。
如上述实施例所述，本公开实施例提供的图像处理方法可以利用神经网络实现，如可以利用图6示出的网络结构实现，下面对训练神经网络的过程进行说明。
首先可以得到训练图像组，训练图像中可以包括至少两个人物对象的虹膜图像，每个人物对象的虹膜图像至少为一个，而且每个虹膜图像的分辨率、图像质量、大小均可以不同，从而可以提高神经网络的适用性。
而后可以利用神经网络执行训练图像的图像处理，得到每个训练图像经图像处理得到的各特征分组对应的分组特征图。进而基于得到的分组特征图得到神经网络的网络损失，在该网络损失小于损失阈值的情况下，可以表示神经网络的检测精度达到要求，可以进行应用，在网络损失大于或者等于损失阈值的情况下，可以反馈调节神经网网络的参数，如卷积参数等，直至得到的损失函数小于损失阈值。其中损失阈值可以为根据需求设定的值，如可以为0.1，但不作为本公开的具体限定。
另外，本公开实施例为了提高网络的检测精度，可以根据相同人物对象的虹膜特征图之间的最小关联度，与不同人物对象之间的最大关联度确定网络损失。例如损失函数可以表示为：
Figure PCTCN2019121695-appb-000002
Figure PCTCN2019121695-appb-000003
L total＝λ 1L 1+λ 1L 2。
其中，Ls表示同一特征融合分支得到的分组特征图对应的网络损失，P表示人物对象的总数，K表示每个人物对象的虹膜图像的总数，s表示特征分组的组数，m表示共同的虹膜区域，B表示分组特征图的列数，
Figure PCTCN2019121695-appb-000004
表示第s分组中第i个人物对象的第a个虹膜图像的特征，
Figure PCTCN2019121695-appb-000005
表示第s分组中第j个人物对象的第n个虹膜图像的特征，MMSD函数表示特征之间的关联度，例如，MMSD(f 1,s,f 2,s)表示对两个训练图像中的一个训练图像的分组特征图f 1,s进行列转置后得到的特征图
Figure PCTCN2019121695-appb-000006
与第二个训练图像的分组特征图f 2,s之间的关联度的最小值。L total表示针对不同特征融合分支得到的分组特征图对应的网络损失的加权和，即整个神经网络的网络损失，λ 1和λ 1分别表示加权系数，L 1和L 2分别表示两个分组对应的网络损失。
在本公开实施例中，通过对虹膜图像执行预处理，对虹膜图像中的虹膜区域进行定位和分割，得到虹膜位置与虹膜区域对应的分割结果，同时还可以对虹膜图像执行多尺度特征提取和多尺度特征融合，获得高精度的虹膜特征图，而后利用分割结果与虹膜特征图执行虹膜图像的身份识别，确定各虹膜图像是否对应于同一对象。通过上述配置，可以通过多尺度特征提取和多尺度融合特征的方式使得提取的底层特征以及高层特征充分的融合，使得最后得到的虹膜特征兼顾底层的纹理特征和高层的分类特征，提高特征提取精度，同时还可以利用分割结果和虹膜特征图的结合，仅考虑虹膜区域的特征部分，减少其他区域的影响，更加精确的识别出虹膜图像是否对应于同一对象，检测结果较高。另外，本公开实施例可以针对虹膜图像中纹理区域重要性不同的特点，在神经网络中采用空间注意力机制，让网络自适应的学习虹膜特征。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。
此外，本公开还提供了图像处理装置、电子设备、计算机可读存储介质、程序，上述均可用来实现本公开提供的任一种图像处理方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。
图9示出根据本公开实施例的一种图像处理装置的框图；，如图9所示，所述图像处理装置包括：
获取模块10，用于获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；
检测模块20，用于检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；
特征处理模块30，用于对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；
比对模块40，用于利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述待比对的两个虹膜图像是否对应于同一对象。
在一些可能的实施方式中，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述特征处理模块还用于对每个所述分组特征对应的所述分组特征图执行加权和处理，得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述分割结果包括所述虹膜图像中虹膜区域对应的掩码图，所述掩码图中的第一标识表示所述虹膜区域，所述掩码图中的第二标识表示所述虹膜区域以外的位置区域。
在一些可能的实施方式中，所述检测二模块还用于对所述虹膜图像执行目标检测处理，确定所述虹膜图像的所述虹膜位置以及瞳孔位置；
基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图像中虹膜区域的分割结果。
在一些可能的实施方式中，所述检测模块还用于分别对所述虹膜图像的虹膜位置对应的图像区域和所述分割结果执行归一化处理；
所述特征处理模块还用于：
对归一化处理后的所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图。
在一些可能的实施方式中，所述比对模块还用于利用所述至少两个虹膜图像分别对应的分割结果，确定所述至少两个虹膜图像中同为虹膜区域的第一位置；
分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；
根据所述至少两个虹膜图像分别对应的第四子特征图之间的关联度，确定所述至少两个虹膜图像的比对结果。
在一些可能的实施方式中，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，确定所述至少两个虹膜图像对应于同一对象。
在一些可能的实施方式中，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，确定所述至少两个虹膜图像对应于不同对象。
在一些可能的实施方式中，所述装置包括神经网络，所述神经网络包括所述获取模块、所述检测模块、所述特征处理模块以及所述比对模块。
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述。
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为上述方法。
电子设备可以被提供为终端、服务器或其它形态的设备。
图10示出根据本公开实施例的一种电子设备的框图。例如，电子设备800可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。
参照图10，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
本公开实施例还提供了一种计算机程序产品，包括计算机可读代码，当计算机可读代码在设备上运行时，设备中的处理器执行用于实现如上任一实施例提供的方法的指令。
该计算机程序产品可以具体通过硬件、软件或其结合的方式实现。在一个可选实施例中，所述计算机程序产品具体体现为计算机存储介质，在另一个可选实施例中，计算机程序产品具体体现为软件产品，例如软件开发包(Software Development Kit，SDK)等等。
图11示出根据本公开实施例的另一种电子设备的框图。例如，电子设备1900可以被提供为一服务器。参照图11，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM或类似。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/ 或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims (25)

  1. 一种图像处理方法，其特征在于，包括：
    获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；
    检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；
    对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；
    利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理得到的比对结果确定所述至少两个虹膜图像是否对应于同一对象。
  2. 根据权利要求1所述的方法，其特征在于，所述对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图，包括：
    对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
    利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
    基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
    基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
  3. 根据权利要求2所述的方法，其特征在于，所述基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图，包括：
    对所述特征分组中的所述至少两个尺度的特征图的连接特征图执行第一卷积处理，得到第一子特征图；
    对所述第一子特征图执行第二卷积处理以及激活函数处理，得到第二子特征图，所述第二子特征图表示所述第一子特征图对应的注意力系数；
    将所述第一子特征图和所述第二子特征图的乘积结果与所述第一子特征图相加，得到第三子特征图；
    对所述第三子特征图执行第三卷积处理，得到所述特征分组对应的分组特征图。
  4. 根据权利要求2或3所述的方法，其特征在于，所述基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图，包括：
    对每个所述分组特征对应的所述分组特征图执行加权和处理，得到所述虹膜图像对应的虹膜特征图。
  5. 根据权利要求1-4中任意一项所述的方法，其特征在于，所述分割结果包括所述虹膜图像中虹膜区域对应的掩码图，所述掩码图中的第一标识表示所述虹膜区域，所述掩码图中的第二标识表示所述虹膜区域以外的位置区域。
  6. 根据权利要求1-5中任意一项所述的方法，其特征在于，所述检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果，包括：
    对所述虹膜图像执行目标检测处理，确定所述虹膜图像的虹膜位置以及瞳孔位置；
    基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图像中虹膜区域的分割结果。
  7. 根据权利要求1-6中任意一项所述的方法，其特征在于，所述检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果，还包括：
    分别对所述虹膜图像的所述虹膜位置对应的图像区域和所述分割结果执行归一化处理；
    所述对所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图，还包括：
    对归一化处理后的所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图。
  8. 根据权利要求1-7中任意一项所述的方法，其特征在于，利用所述至少两个虹膜图像分别对应的 所述分割结果和所述虹膜特征图，执行比对处理，包括：
    利用所述至少两个虹膜图像分别对应的所述分割结果，确定所述至少两个虹膜图像中同为虹膜区域的第一位置；
    分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；
    根据所述至少两个虹膜图像分别对应的所述第四子特征图之间的关联度，对所述至少两个虹膜图像执行比对处理。
  9. 根据权利要求8所述的方法，其特征在于，所述基于所述比对处理得到的比对结果确定所述至少两个虹膜图像是否对应于同一对象，包括：
    在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，确定所述至少两个虹膜图像对应于同一对象。
  10. 根据权利要求8或9所述的方法，其特征在于，所述基于比对结果确定所述至少两个虹膜图像是否对应于同一对象，还包括：
    在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，确定所述至少两个虹膜图像对应于不同对象。
  11. 根据权利要求1-10中任意一项所述的方法，其特征在于，所述图像处理方法通过卷积神经网络实现。
  12. 一种图像处理装置，其特征在于，包括：
    获取模块，用于获取虹膜图像组，所述虹膜图像组包括待比对的至少两个虹膜图像；
    检测模块，用于检测所述虹膜图像中的虹膜位置，以及所述虹膜图像中虹膜区域的分割结果；
    特征处理模块，用于对所述虹膜位置对应的图像区域执行多尺度特征提取和多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图；
    比对模块，用于利用所述至少两个虹膜图像分别对应的所述分割结果和所述虹膜特征图，执行比对处理，基于所述比对处理的比对结果确定所述至少两个虹膜图像是否对应于同一对象。
  13. 根据权利要求12所述的装置，其特征在于，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
    利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
    基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
    基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
  14. 根据权利要求13所述的装置，其特征在于，所述特征处理模块还用于对所述虹膜图像中所述虹膜位置对应的图像区域执行所述多尺度特征提取处理，得到多个尺度的特征图；
    利用所述多个尺度的特征图，形成至少一个特征分组，所述特征分组包括所述多个尺度特征图中至少两个尺度的特征图；
    基于注意力机制，对所述特征分组中的特征图执行所述多尺度特征融合处理，得到所述特征分组对应的分组特征图；
    基于所述特征分组对应的分组特征图得到所述虹膜图像对应的虹膜特征图。
  15. 根据权利要求13或14所述的装置，其特征在于，所述特征处理模块还用于对每个所述分组特征对应的所述分组特征图执行加权和处理，得到所述虹膜图像对应的虹膜特征图。
  16. 根据权利要求12-15中任意一项所述的装置，其特征在于，所述分割结果包括所述虹膜图像中虹膜区域对应的掩码图，所述掩码图中的第一标识表示所述虹膜区域，所述掩码图中的第二标识表示所述虹膜区域以外的位置区域。
  17. 根据权利要求12-16中任意一项所述的装置，其特征在于，所述检测模块还用于对所述虹膜图像执行目标检测处理，确定所述虹膜图像的所述虹膜位置以及瞳孔位置；
    基于确定的所述虹膜位置和所述瞳孔位置，对所述虹膜图像执行所述分割处理，得到所述虹膜图 像中虹膜区域的分割结果。
  18. 根据权利要求12-17中任意一项所述的装置，其特征在于，所述检测模块还用于分别对所述虹膜图像的虹膜位置对应的图像区域和所述分割结果执行归一化处理；
    所述特征处理模块还用于：对归一化处理后的所述虹膜位置对应的图像区域执行所述多尺度特征提取和所述多尺度特征融合处理，得到所述虹膜图像对应的虹膜特征图。
  19. 根据权利要求12-18中任意一项所述的装置，其特征在于，所述比对模块还用于利用所述至少两个虹膜图像分别对应的分割结果，确定所述至少两个虹膜图像中同为虹膜区域的第一位置；
    分别确定所述至少两个虹膜图像的虹膜特征图中与所述第一位置对应的第四子特征图；
    根据所述至少两个虹膜图像分别对应的第四子特征图之间的关联度，确定所述至少两个虹膜图像的比对结果。
  20. 根据权利要求19所述的装置，其特征在于，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度大于第一阈值的情况下，确定所述至少两个虹膜图像对应于同一对象。
  21. 根据权利要求19或20所述的装置，其特征在于，所述比对模块还用于在所述至少两个虹膜图像分别对应的第四子特征图之间的关联度小于或者等于第一阈值的情况下，确定所述至少两个虹膜图像对应于不同对象。
  22. 根据权利要求12-21中任意一项所述的装置，其特征在于，所述装置包括神经网络，所述神经网络包括所述获取模块、所述检测模块、所述特征处理模块以及所述比对模块。
  23. 一种电子设备，其特征在于，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器被配置为调用所述存储器存储的指令，以执行权利要求1至11中任意一项所述的方法。
  24. 一种计算机可读存储介质，其上存储有计算机程序指令，其特征在于，所述计算机程序指令被处理器执行时实现权利要求1至11中任意一项所述的方法。
  25. 一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1-11中的任一权利要求所述的方法。
PCT/CN2019/121695 2019-09-26 2019-11-28 图像处理方法及装置、电子设备和存储介质 WO2021056808A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
KR1020217008623A KR20210047336A (ko) 2019-09-26 2019-11-28 이미지 처리 방법 및 장치, 전자 기기 및 기억 매체
SG11202013254VA SG11202013254VA (en) 2019-09-26 2019-11-28 Image processing method and device, electronic apparatus and storage medium
JP2021500196A JP7089106B2 (ja) 2019-09-26 2019-11-28 画像処理方法及び装置、電子機器、コンピュータ読取可能記憶媒体及びコンピュータプログラム
US17/137,819 US11532180B2 (en) 2019-09-26 2020-12-30 Image processing method and device and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910919121.9A CN110688951B (zh) 2019-09-26 2019-09-26 图像处理方法及装置、电子设备和存储介质
CN201910919121.9 2019-09-26

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/137,819 Continuation US11532180B2 (en) 2019-09-26 2020-12-30 Image processing method and device and storage medium

Publications (1)

Publication Number Publication Date
WO2021056808A1 true WO2021056808A1 (zh) 2021-04-01

Family

ID=69110423

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/121695 WO2021056808A1 (zh) 2019-09-26 2019-11-28 图像处理方法及装置、电子设备和存储介质

Country Status (7)

Country Link
US (1) US11532180B2 (zh)
JP (1) JP7089106B2 (zh)
KR (1) KR20210047336A (zh)
CN (1) CN110688951B (zh)
SG (1) SG11202013254VA (zh)
TW (1) TWI724736B (zh)
WO (1) WO2021056808A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113642393A (zh) * 2021-07-07 2021-11-12 重庆邮电大学 基于注意力机制的多特征融合视线估计方法
CN113643305A (zh) * 2021-08-10 2021-11-12 珠海复旦创新研究院 一种基于深度网络上下文提升的人像检测与分割方法
CN114519723A (zh) * 2021-12-24 2022-05-20 上海海洋大学 一种基于金字塔影像分割的陨石坑自动提取方法
WO2023169582A1 (zh) * 2022-03-11 2023-09-14 北京字跳网络技术有限公司 图像增强方法、装置、设备及介质

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111292343B (zh) * 2020-01-15 2023-04-28 东北大学 一种基于多视角下的肺叶分割方法和装置
CN111260627B (zh) * 2020-01-15 2023-04-28 东北大学 一种基于肺叶的肺气肿区域判断方法和装置
CN111401145B (zh) * 2020-02-26 2022-05-03 三峡大学 一种基于深度学习与ds证据理论的可见光虹膜识别方法
CN111858989B (zh) * 2020-06-09 2023-11-10 西安工程大学 一种基于注意力机制的脉冲卷积神经网络的图像分类方法
CN111681273B (zh) * 2020-06-10 2023-02-03 创新奇智(青岛)科技有限公司 图像分割方法、装置、电子设备及可读存储介质
CN111862034B (zh) * 2020-07-15 2023-06-30 平安科技（深圳）有限公司 图像检测方法、装置、电子设备及介质
CN112184635A (zh) * 2020-09-10 2021-01-05 上海商汤智能科技有限公司 目标检测方法、装置、存储介质及设备
CN112288723B (zh) * 2020-10-30 2023-05-23 北京市商汤科技开发有限公司 缺陷检测方法、装置、计算机设备及存储介质
CN112287872B (zh) * 2020-11-12 2022-03-25 北京建筑大学 基于多任务神经网络的虹膜图像分割、定位和归一化方法
CN113076926B (zh) * 2021-04-25 2022-11-18 华南理工大学 一种带语义引导的多尺度目标检测方法及系统
CN113486815B (zh) * 2021-07-09 2022-10-21 山东力聚机器人科技股份有限公司 一种行人重识别系统和方法、计算机设备及存储介质
CN113591843B (zh) * 2021-07-12 2024-04-09 中国兵器工业计算机应用技术研究所 仿初级视觉皮层的目标检测方法、装置及设备
CN114359120B (zh) * 2022-03-21 2022-06-21 深圳市华付信息技术有限公司 遥感影像处理方法、装置、设备及存储介质
CN114998980B (zh) * 2022-06-13 2023-03-31 北京万里红科技有限公司 一种虹膜检测方法、装置、电子设备及存储介质
CN115100730B (zh) * 2022-07-21 2023-08-08 北京万里红科技有限公司 虹膜活体检测模型的训练方法、虹膜活体检测方法及装置
CN116704666A (zh) * 2023-06-21 2023-09-05 合肥中科类脑智能技术有限公司 售卖方法及计算机可读存储介质、自动售卖机

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106778664A (zh) * 2016-12-29 2017-05-31 天津中科智能识别产业技术研究院有限公司 一种虹膜图像中虹膜区域的分割方法及其装置
CN107506754A (zh) * 2017-09-19 2017-12-22 厦门中控智慧信息技术有限公司 虹膜识别方法、装置及终端设备
CN108229531A (zh) * 2017-09-29 2018-06-29 北京市商汤科技开发有限公司 对象特征处理方法、装置、存储介质和电子设备
US20180218213A1 (en) * 2017-02-02 2018-08-02 Samsung Electronics Co., Ltd. Device and method of recognizing iris
CN109426770A (zh) * 2017-08-24 2019-03-05 合肥虹慧达科技有限公司 虹膜识别方法

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6247813B1 (en) * 1999-04-09 2001-06-19 Iritech, Inc. Iris identification system and method of identifying a person through iris recognition
JP3586431B2 (ja) * 2001-02-28 2004-11-10 松下電器産業株式会社 個人認証方法および装置
JP2004206444A (ja) * 2002-12-25 2004-07-22 Matsushita Electric Ind Co Ltd 個人認証方法および虹彩認証装置
CN100498837C (zh) * 2004-05-10 2009-06-10 松下电器产业株式会社 虹彩注册方法、虹彩注册装置
KR100629550B1 (ko) * 2004-11-22 2006-09-27 아이리텍 잉크 다중스케일 가변영역분할 홍채인식 방법 및 시스템
CN101539990B (zh) * 2008-03-20 2011-05-11 中国科学院自动化研究所 一种虹膜图像鲁棒特征选择和快速比对的方法
CN102844766B (zh) * 2011-04-20 2014-12-24 中国科学院自动化研究所 基于人眼图像的多特征融合身份识别方法
CN104063872B (zh) * 2014-07-04 2017-02-15 西安电子科技大学 基于改进视觉注意模型的序列图像显著区域检测方法
CN106326841A (zh) * 2016-08-12 2017-01-11 合肥虹视信息工程有限公司 一种快速虹膜识别算法
RU2670798C9 (ru) * 2017-11-24 2018-11-26 Самсунг Электроникс Ко., Лтд. Способ аутентификации пользователя по радужной оболочке глаз и соответствующее устройство
CN110059589B (zh) * 2019-03-21 2020-12-29 昆山杜克大学 一种基于Mask R-CNN神经网络的虹膜图像中虹膜区域的分割方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106778664A (zh) * 2016-12-29 2017-05-31 天津中科智能识别产业技术研究院有限公司 一种虹膜图像中虹膜区域的分割方法及其装置
US20180218213A1 (en) * 2017-02-02 2018-08-02 Samsung Electronics Co., Ltd. Device and method of recognizing iris
CN109426770A (zh) * 2017-08-24 2019-03-05 合肥虹慧达科技有限公司 虹膜识别方法
CN107506754A (zh) * 2017-09-19 2017-12-22 厦门中控智慧信息技术有限公司 虹膜识别方法、装置及终端设备
CN108229531A (zh) * 2017-09-29 2018-06-29 北京市商汤科技开发有限公司 对象特征处理方法、装置、存储介质和电子设备

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
YULIN SI ET AL.: "Novel Approaches to Improve Robustness, Accuracy and Rapidity of Iris Recognition Systems", IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, vol. 8, no. 1, 19 January 2012 (2012-01-19), XP011398081, ISSN: 1551-3203 *
ZHAO, YANMING: "THE IRIS RECOGNITION ALGORITHM BASED ON SCALE CORRELATION MULTI-FEATURE EXTRACTION AND FUSION", COMPUTER APPLICATIONS AND SOFTWARE, vol. 30, no. 7, 15 July 2013 (2013-07-15), pages 188 - 192, XP055794560, ISSN: 1000-386X *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113642393A (zh) * 2021-07-07 2021-11-12 重庆邮电大学 基于注意力机制的多特征融合视线估计方法
CN113642393B (zh) * 2021-07-07 2024-03-22 重庆邮电大学 基于注意力机制的多特征融合视线估计方法
CN113643305A (zh) * 2021-08-10 2021-11-12 珠海复旦创新研究院 一种基于深度网络上下文提升的人像检测与分割方法
CN113643305B (zh) * 2021-08-10 2023-08-25 珠海复旦创新研究院 一种基于深度网络上下文提升的人像检测与分割方法
CN114519723A (zh) * 2021-12-24 2022-05-20 上海海洋大学 一种基于金字塔影像分割的陨石坑自动提取方法
CN114519723B (zh) * 2021-12-24 2024-05-28 上海海洋大学 一种基于金字塔影像分割的陨石坑自动提取方法
WO2023169582A1 (zh) * 2022-03-11 2023-09-14 北京字跳网络技术有限公司 图像增强方法、装置、设备及介质

Also Published As

Publication number Publication date
TWI724736B (zh) 2021-04-11
US11532180B2 (en) 2022-12-20
JP2022511217A (ja) 2022-01-31
CN110688951B (zh) 2022-05-31
TW202113756A (zh) 2021-04-01
JP7089106B2 (ja) 2022-06-21
US20210117674A1 (en) 2021-04-22
CN110688951A (zh) 2020-01-14
SG11202013254VA (en) 2021-04-29
KR20210047336A (ko) 2021-04-29

Similar Documents

Publication Publication Date Title
WO2021056808A1 (zh) 图像处理方法及装置、电子设备和存储介质
CN111310616B (zh) 图像处理方法及装置、电子设备和存储介质
WO2021051650A1 (zh) 人脸和人手关联检测方法及装置、电子设备和存储介质
TWI747325B (zh) 目標對象匹配方法及目標對象匹配裝置、電子設備和電腦可讀儲存媒介
CN109522910B (zh) 关键点检测方法及装置、电子设备和存储介质
WO2021031609A1 (zh) 活体检测方法及装置、电子设备和存储介质
US10007841B2 (en) Human face recognition method, apparatus and terminal
US20120321193A1 (en) Method, apparatus, and computer program product for image clustering
WO2021208667A1 (zh) 图像处理方法及装置、电子设备和存储介质
CN109934275B (zh) 图像处理方法及装置、电子设备和存储介质
CN109977860B (zh) 图像处理方法及装置、电子设备和存储介质
CN110532956B (zh) 图像处理方法及装置、电子设备和存储介质
CN111435432B (zh) 网络优化方法及装置、图像处理方法及装置、存储介质
WO2019205605A1 (zh) 人脸特征点的定位方法及装置
CN111259967A (zh) 图像分类及神经网络训练方法、装置、设备及存储介质
US20220270352A1 (en) Methods, apparatuses, devices, storage media and program products for determining performance parameters
TWI770531B (zh) 人臉識別方法、電子設備和儲存介質
WO2023155393A1 (zh) 特征点匹配方法、装置、电子设备、存储介质和计算机程序产品
CN111723715B (zh) 一种视频显著性检测方法及装置、电子设备和存储介质

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2021500196

Country of ref document: JP

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 20217008623

Country of ref document: KR

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19946496

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19946496

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19946496

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 22.05.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 19946496

Country of ref document: EP

Kind code of ref document: A1