CN110689486A - Image processing method, device, equipment and computer storage medium - Google Patents
Image processing method, device, equipment and computer storage medium Download PDFInfo
- Publication number
- CN110689486A CN110689486A CN201810735358.7A CN201810735358A CN110689486A CN 110689486 A CN110689486 A CN 110689486A CN 201810735358 A CN201810735358 A CN 201810735358A CN 110689486 A CN110689486 A CN 110689486A
- Authority
- CN
- China
- Prior art keywords
- illumination
- image
- low
- data set
- normal
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title description 13
- 238000005286 illumination Methods 0.000 claims abstract description 263
- 238000012545 processing Methods 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims description 96
- 238000005070 sampling Methods 0.000 claims description 66
- 238000005282 brightening Methods 0.000 claims description 28
- 238000012937 correction Methods 0.000 claims description 10
- 230000002708 enhancing effect Effects 0.000 description 9
- 241000282414 Homo sapiens Species 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 238000002059 diagnostic imaging Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 241000282412 Homo Species 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G06T5/92—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10052—Images from lightfield camera
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The embodiment of the invention discloses a method, a device and equipment for processing images and a computer storage medium, wherein the method comprises the following steps: obtaining an image request, the image request comprising: a low-light image to be processed; and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed. Therefore, the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can obtain more information from the enhanced low-illumination image.
Description
Technical Field
The embodiment of the invention relates to the technical field of image processing, in particular to an image processing method, device and equipment and a computer storage medium.
Background
Images are the main source of information acquired and exchanged by humans, and therefore, the clarity of images and the proportion of useful information become paramount. For example, a user can easily obtain desired information from a clear image, but if the current image is a low-light image, the user cannot quickly obtain effective information from the image.
Specifically, because of a multi-scale retina enhancement algorithm with color recovery and the like, human eyes generally consider that the brightness of an object perceived by human eyes depends on the illumination of the environment and the reflection of the object surface to the illumination light, so that the low-illumination image can be enhanced by disassembling the image into essence and illumination and enhancing the illumination.
However, when the above method is used for low-light enhancement, parameters are generally set manually, so that the method is not suitable for all low-light conditions, and moreover, the image may have artifacts such as black edges, so that the image quality after the brightness enhancement is poor.
Disclosure of Invention
The embodiment of the invention provides an image processing method, an image processing device, image processing equipment and a computer storage medium, and aims to solve the technical problem that the existing low-illumination image brightening method cannot adapt to all low-illumination conditions.
The first aspect of the present invention provides a method for processing an image, including:
obtaining an image request, the image request comprising: a low-light image to be processed;
and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed.
Another aspect of the present invention provides an image processing apparatus, including:
an image request obtaining module, configured to obtain an image request, where the image request includes: a low-light image to be processed;
and the brightening processing module is used for brightening the low-illumination image to be processed by adopting a pre-configured low-illumination enhancement network according to the image processing request to obtain a normal-illumination image corresponding to the low-illumination image to be processed.
Yet another aspect of the present invention is to provide an image processing apparatus including: a memory, a processor;
a memory; a memory for storing the processor-executable instructions;
wherein the processor is configured to perform the above method by the processor.
Yet another aspect of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the image processing method described above.
The invention provides a method, a device, equipment and a computer storage medium for processing an image, wherein the image request comprises the following steps: a low-light image to be processed; and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed. Therefore, the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can obtain more information from the enhanced low-illumination image.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic flowchart of an image processing method according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an image processing apparatus according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of an image processing apparatus according to a third embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flowchart of an image processing method according to an embodiment of the present invention, as shown in fig. 1, the method includes:
step 101, obtaining an image request, wherein the image request comprises: a low-light image to be processed;
and step 102, according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed.
In practical applications, images are the main sources of information acquired and exchanged by human beings, and are applied to the fields of security monitoring, military reconnaissance, medical imaging, machine vision and the like. However, under the low-light condition, the color information of the digital image acquired by the image acquisition device often hardly reflects the intrinsic color of the imaged object, because the acquisition device has no color constancy of the human visual system, and cannot peel off the influence caused by external low light and shooting angle, resulting in color loss and blurred texture of the generated low-light color image, and further resulting in that corresponding effective information cannot be acquired from the low-light image. Therefore, to achieve enhancement of the low-light image, an image request may first be obtained, wherein the image request includes the low-light image to be processed. And according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed and obtain a normal-illumination image corresponding to the low-illumination image to be processed, so that a user can extract effective information from the normal-illumination image. The low-illumination enhancement network is obtained in advance according to training of a plurality of low-illumination images and normal-illumination images and is used for effectively enhancing the low-illumination images.
In the method for processing an image provided in this embodiment, an image request is obtained, where the image request includes: a low-light image to be processed; and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed. Therefore, the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can obtain more information from the enhanced low-illumination image.
Further, on the basis of the above embodiment, the method further includes:
acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
building a convolution network with a global illumination view;
and training the convolution network according to the training data set to obtain the low-illumination enhancement network.
In this embodiment, in order to achieve effective enhancement of the low-light image, training of the low-light network is first required. Specifically, a training data set may be obtained, where the training data set includes a plurality of normal-illumination sample images and their corresponding low-illumination sample images. And a convolution network with a global illumination view field is established, and it needs to be explained that the convolution network with the global illumination view field can analyze the global illumination distribution of the current low-illumination image, and in addition, the detail restoration can be carried out on the low-illumination image, so that the adjustment of the details in the low-illumination image can be realized on the basis of realizing the enhancement of the low-illumination image, and the quality of the enhanced image can be further improved. Further, a convolutional network with a global illumination field of view is trained to obtain a low-illumination enhancement network based on the training dataset.
According to the image processing method provided by the embodiment, a convolutional network with a global illumination field is established by acquiring a training data set comprising a plurality of normal-illumination sample images and corresponding low-illumination sample images in advance, and the convolutional network is trained through data in the training data set to obtain a low-illumination enhancement network, so that the details in the low-illumination images can be adjusted on the basis of enhancing the low-illumination images, and the quality of the enhanced images can be improved.
Further, on the basis of any of the above embodiments, the method comprises:
collecting a sample image of normal illumination;
respectively carrying out low-light processing on the collected normal-illumination sample images to obtain corresponding low-illumination sample images; combining each sample image with normal illumination and the corresponding sample image with low illumination into a picture pair to obtain the training data set;
building a convolution network with a global illumination view;
and training the convolution network according to the training data set to obtain the low-illumination enhancement network.
In this embodiment, in order to effectively enhance the low-illumination image, a training data set needs to be constructed first, specifically, a sample image of normal illumination may be collected, low-illumination processing may be performed on the collected sample image of normal illumination respectively, and specifically, parameters of the sample image of normal illumination may be adjusted randomly to obtain a low-illumination sample image corresponding to the sample image of normal illumination. Further, each sample image with normal illumination and the corresponding sample image with low illumination are combined into a picture pair respectively to obtain a training data set, a convolution network with a global illumination field of view is built, and the convolution network is trained through the data in the training data set to obtain the low illumination enhancement network.
For example, in practical applications, 1000 normal illumination images in the original image file format can be collected in advance, the exposure parameter is randomly adjusted to be between-5 and 0, the vividness parameter is adjusted to be between-100 and 0, and the contrast parameter is adjusted to be between-100 and 0, so that the low illumination image in the original image file format is generated. Both the normal-light picture and the low-light picture are converted into an 8-bit picture format with a resolution of 400 × 600. And combining the corresponding normal light picture Y and the low light picture X into a picture pair to generate a training data set. And constructing a convolution network with a global illumination view, and training the convolution network through the data in the training data set to obtain the low-illumination enhancement network.
In the image processing method provided by this embodiment, a sample image under normal illumination is obtained in advance, and low-light processing is performed on the sample image under normal illumination at random, so as to obtain a corresponding low-light sample image; and combining each sample image of normal illumination and the corresponding sample image of low illumination into a picture pair to obtain the training data set, thereby providing a basis for enhancing the low illumination image.
Further, on the basis of any of the above embodiments, the method comprises:
acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
performing global illumination distribution estimation processing on an input low-illumination image to obtain an up-sampling output result and a down-sampling output result;
combining the up-sampling output result and the down-sampling output result to obtain a combined result;
carrying out convolution and linear correction processing on the merged result for three times in sequence to obtain an output normal illumination image and a corresponding brightening result;
and training the convolution network according to the training data set to obtain the low-illumination enhancement network.
In this embodiment, after the training data set is obtained, global illumination estimation processing is performed on an input low-illumination image, an up-sampling output result and a down-sampling output result are obtained, the up-sampling output result and the down-sampling output result are combined, a combined result is obtained, and the combined result is sequentially subjected to triple convolution and linear correction processing to obtain an output normal illumination image and a corresponding brightening result thereof, so that a convolution network with a global illumination field is obtained. It should be noted that the convolution network with the global illumination field can analyze the global illumination distribution of the current low-illumination image, and can also perform detail restoration on the low-illumination image, so that the details in the low-illumination image can be adjusted on the basis of enhancing the low-illumination image, and the quality of the enhanced image can be further improved.
For practical applications, for example, the input low-light picture X is passed through a down-sampling module D0And obtaining the low-resolution feature f. D0The method comprises a nearest neighbor operation, a convolution operation and a modified linear unit. D0The nearest neighbor scaling operation of (2) scales the picture to a fixed 96 x 96 resolution. The low resolution feature f is then passed through an auto-encoder. The self-encoder is down-sampled by 5 symmetrical groups D1，D2，...，D5And an up-sampling module U1，U2，...，U5And (4) forming. Each of the down-sampling module and the up-sampling module consists of a nearest neighbor operation, a convolution operation and a modified linear unit.The nearest neighbor operation of the down-sampling module is to reduce the input to 0.5 times of the original input, and the nearest neighbor operation of the up-sampling module is to enlarge the input to 2 times of the original input. Each down-sampling module D1，D2，...，D5And an up-sampling module U5The input of (1) is the output of the previous layer. For the up-sampling module Ut(t ═ 1,2,. 4), the symmetric down-sampled layer D is comparedtAnd the last up-sampling module Ut+1The output results are added to obtain an up-sampling module UtIs input. Finally, U5Is passed through an up-sampling module U0The result of the first step, i.e. the estimation of the global illumination distribution, is obtained. Wherein, the up-sampling module U0The nearest neighbor operation of (2) scales the input to the same resolution as input picture X. Further, an up-sampling module U0Output result and down sampling module D0The output results are merged and pass through three convolution modules to obtain the final brightening result
Wherein each convolution module consists of a convolution operation and a modified linear unit.
In the image processing method provided by this embodiment, global illumination distribution estimation processing is performed on an input low-illumination image to obtain an up-sampling output result and a down-sampling output result; combining the up-sampling output result and the down-sampling output result to obtain a combined result; and sequentially carrying out convolution and linear correction processing on the merged result for three times to obtain the output normal illumination image and the corresponding brightening result, thereby obtaining a convolution network with a global illumination view, realizing the adjustment of details in the low illumination image on the basis of realizing the enhancement of the low illumination image, and further improving the quality of the enhanced image.
Further, on the basis of any of the above embodiments, the method further includes:
acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
performing global illumination distribution estimation processing on an input low-illumination image to obtain an up-sampling output result and a down-sampling output result;
combining the up-sampling output result and the down-sampling output result to obtain a combined result;
carrying out convolution and linear correction processing on the merged result for three times in sequence to obtain an output normal illumination image and a corresponding brightening result;
and training the convolution network according to the training data set, the output normal light image and the corresponding brightening result thereof to obtain the low-light enhancement network.
In this embodiment, after the training data set is acquired and the convolutional network with the global illumination field is established, the convolutional network may be trained according to the training data set, the output normal illumination image and the corresponding brightening result thereof to acquire the low-illumination enhancement network.
According to the image processing method provided by the embodiment, the convolutional network is trained to acquire the low-illumination enhancement network according to the training data set, the output normal-illumination image and the corresponding brightening result, so that the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can acquire more information from the enhanced low-illumination image.
Fig. 2 is a schematic structural diagram of an image processing apparatus according to a second embodiment of the present invention, as shown in fig. 2, the apparatus includes:
an image request obtaining module 21, configured to obtain an image request, where the image request includes: a low-light image to be processed;
and the brightening processing module 22 is configured to brighten the low-illumination image to be processed by using a pre-configured low-illumination enhancement network according to the image processing request, and obtain a normal-illumination image corresponding to the low-illumination image to be processed.
In practical applications, images are the main sources of information acquired and exchanged by human beings, and are applied to the fields of security monitoring, military reconnaissance, medical imaging, machine vision and the like. However, under the low-light condition, the color information of the digital image acquired by the image acquisition device often hardly reflects the intrinsic color of the imaged object, because the acquisition device has no color constancy of the human visual system, and cannot peel off the influence caused by external low light and shooting angle, resulting in color loss and blurred texture of the generated low-light color image, and further resulting in that corresponding effective information cannot be acquired from the low-light image. Therefore, in order to realize the enhancement of the low-light image, first, the image request obtaining module 21 may obtain an image request, where the image request includes the low-light image to be processed. The brightening processing module 22 adopts a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed according to the image processing request, and obtains a normal-illumination image corresponding to the low-illumination image to be processed, so that the user can extract effective information from the normal-illumination image. The low-illumination enhancement network is obtained in advance according to training of a plurality of low-illumination images and normal-illumination images and is used for effectively enhancing the low-illumination images.
The image processing apparatus provided in this embodiment, by obtaining an image request, the image request includes: a low-light image to be processed; and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed. Therefore, the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can obtain more information from the enhanced low-illumination image.
Further, on the basis of the above embodiment, the apparatus further includes:
the training data set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
the convolution network building module is used for building a convolution network with a global illumination view;
and the training module is used for training the convolution network according to the training data set so as to obtain the low-illumination enhancement network.
In this embodiment, in order to achieve effective enhancement of the low-light image, training of the low-light network is first required. Specifically, the training data set obtaining module may obtain a training data set, where the training data set includes a plurality of normal-illumination sample images and low-illumination sample images corresponding to the normal-illumination sample images. The convolution network with the global illumination field of view is constructed by the convolution network construction module, and it needs to be explained that the convolution network with the global illumination field of view can analyze the global illumination distribution of the current low-illumination image, and in addition, can also carry out detail restoration on the low-illumination image, so that the adjustment of details in the low-illumination image can be realized on the basis of realizing the enhancement of the low-illumination image, and the quality of the enhanced image can be further improved. Further, the training module trains the convolutional network with the global illumination field of view according to the training data set to obtain a low-illumination enhancement network.
The image processing device provided by this embodiment builds a convolution network with a global illumination field by acquiring a training data set including a plurality of normal-illumination sample images and corresponding low-illumination sample images in advance, and trains the convolution network through data in the training data set to obtain a low-illumination enhancement network, so that details in the low-illumination image can be adjusted on the basis of enhancing the low-illumination image, and the quality of the enhanced image can be improved.
Further, on the basis of any of the above embodiments, the apparatus comprises:
the training data set acquisition module specifically comprises:
the sample image acquisition unit is used for acquiring a sample image under normal illumination;
the low light processing unit is used for respectively carrying out low light processing on the collected sample images with normal illumination to obtain corresponding sample images with low illumination; combining each sample image with normal illumination and the corresponding sample image with low illumination into a picture pair to obtain the training data set;
the convolution network building module is used for building a convolution network with a global illumination view;
and the training module is used for training the convolution network according to the training data set so as to obtain the low-illumination enhancement network.
In this embodiment, in order to effectively enhance the low-light image, a training data set needs to be constructed first, specifically, the sample image collecting unit may collect a sample image of normal illumination, the low-light processing unit performs low-light processing on the collected sample image of normal illumination, and specifically, parameters of the sample image of normal illumination may be adjusted randomly to obtain a low-light sample image corresponding to the sample image of normal illumination. Further, each sample image with normal illumination and the corresponding sample image with low illumination are combined into a picture pair respectively to obtain a training data set, a convolution network with a global illumination field of view is built, and the convolution network is trained through the data in the training data set to obtain the low illumination enhancement network.
For example, in practical applications, 1000 normal illumination images in the original image file format can be collected in advance, the exposure parameter is randomly adjusted to be between-5 and 0, the vividness parameter is adjusted to be between-100 and 0, and the contrast parameter is adjusted to be between-100 and 0, so that the low illumination image in the original image file format is generated. Both the normal-light picture and the low-light picture are converted into an 8-bit picture format with a resolution of 400 × 600. And combining the corresponding normal light picture Y and the low light picture X into a picture pair to generate a training data set. And constructing a convolution network with a global illumination view, and training the convolution network through the data in the training data set to obtain the low-illumination enhancement network.
The image processing device provided in this embodiment obtains a sample image under normal illumination in advance, and performs low-light processing on the sample image under normal illumination at random to obtain a corresponding low-light sample image; and combining each sample image of normal illumination and the corresponding sample image of low illumination into a picture pair to obtain the training data set, thereby providing a basis for enhancing the low illumination image.
Further, on the basis of any of the above embodiments, the apparatus comprises:
the training data set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
the convolutional network building module specifically comprises:
the estimating unit is used for carrying out global illumination distribution estimation processing on the input low-illumination image to obtain an up-sampling output result and a down-sampling output result;
the merging unit is used for merging the up-sampling output result and the down-sampling output result to obtain a merging result;
the brightening unit is used for sequentially carrying out three times of convolution and linear correction processing on the combined result so as to obtain an output normal illumination image and a brightening result corresponding to the output normal illumination image;
and the training module is used for training the convolution network according to the training data set so as to obtain the low-illumination enhancement network.
In this embodiment, after the training data set is obtained, the estimation unit performs global illumination estimation processing on the input low-illumination image to obtain an up-sampling output result and a down-sampling output result, the merging unit performs merging processing on the up-sampling output result and the down-sampling output result to obtain a merged result, and the brightening unit performs triple convolution and linear correction processing on the merged result in sequence to obtain an output normal-illumination image and a brightening result corresponding to the output normal-illumination image, so as to obtain a convolution network with a global illumination view. It should be noted that the convolution network with the global illumination field can analyze the global illumination distribution of the current low-illumination image, and can also perform detail restoration on the low-illumination image, so that the details in the low-illumination image can be adjusted on the basis of enhancing the low-illumination image, and the quality of the enhanced image can be further improved.
For practical applications, for example, the input low-light picture X is passed through a down-sampling module D0And obtaining the low-resolution feature f. D0The method comprises a nearest neighbor operation, a convolution operation and a modified linear unit. D0The nearest neighbor scaling operation of (2) scales the picture to a fixed 96 x 96 resolution. The low resolution feature f is then passed through an auto-encoder. The self-encoder is down-sampled by 5 symmetrical groups D1，D2，...，D5And an up-sampling module U1，U2，...，U5And (4) forming. Each of the down-sampling module and the up-sampling module consists of a nearest neighbor operation, a convolution operation and a modified linear unit. The nearest neighbor operation of the down-sampling module is to reduce the input to 0.5 times of the original input, and the nearest neighbor operation of the up-sampling module is to enlarge the input to 2 times of the original input. Each down-sampling module D1，D2，...，D5And an up-sampling module U5The input of (1) is the output of the previous layer. For the up-sampling module Ut(t ═ 1,2,. 4), the symmetric down-sampled layer D is comparedtAnd the last up-sampling module Ut+1The output results are added to obtain an up-sampling module UtIs input. Finally, U5Is passed through an up-sampling module U0The result of the first step, i.e. the estimation of the global illumination distribution, is obtained. Wherein, the up-sampling module U0The nearest neighbor operation of (2) scales the input to the same resolution as input picture X. Further, an up-sampling module U0Output result and down sampling module D0The output results are merged, and the final brightening result is obtained through a module consisting of three convolution operations and a correction linear unit
The image processing apparatus provided in this embodiment obtains an up-sampling output result and a down-sampling output result by performing global illumination distribution estimation processing on an input low-illumination image; combining the up-sampling output result and the down-sampling output result to obtain a combined result; and sequentially carrying out convolution and linear correction processing on the merged result for three times to obtain the output normal illumination image and the corresponding brightening result, thereby obtaining a convolution network with a global illumination view, realizing the adjustment of details in the low illumination image on the basis of realizing the enhancement of the low illumination image, and further improving the quality of the enhanced image.
Further, on the basis of any one of the above embodiments, the apparatus further includes:
the training data set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
the convolutional network building module specifically comprises:
the estimating unit is used for carrying out global illumination distribution estimation processing on the input low-illumination image to obtain an up-sampling output result and a down-sampling output result;
combining the up-sampling output result and the down-sampling output result to obtain a combined result;
the brightening unit is used for sequentially carrying out three times of convolution and linear correction processing on the combined result so as to obtain an output normal illumination image and a brightening result corresponding to the output normal illumination image;
the training module specifically comprises:
and the training unit is used for training the convolution network according to the training data set, the output normal light image and the corresponding brightening result thereof so as to obtain the low-illumination enhancement network.
In this embodiment, after the training data set is acquired and the convolutional network with the global illumination field is established, the training unit may train the convolutional network according to the training data set, the output normal illumination image and the corresponding brightening result thereof to acquire the low-illumination enhancement network.
According to the image processing device provided by the embodiment, the convolutional network is trained to acquire the low-illumination enhancement network according to the training data set, the output normal-illumination image and the corresponding brightening result, so that the enhancement of the low-illumination image to be processed can be accurately realized, and further, a user can acquire more information from the enhanced low-illumination image.
Fig. 3 is a schematic structural diagram of an image processing apparatus according to a third embodiment of the present invention, as shown in fig. 3, the image processing apparatus includes a memory 31, a processor 32;
a memory 31; a memory 31 for storing instructions executable by the processor 32;
wherein the processor 32 is configured to perform the above-described method by the processor 32.
When at least a part of the functions of the image processing method in the embodiment of the present invention are implemented by software, the embodiment of the present invention further provides a computer storage medium for storing computer software instructions for the above-mentioned image processing, which, when executed on a computer, enable the computer to perform various possible image processing methods in the above-mentioned method embodiments. The processes or functions described in accordance with the embodiments of the present invention may be generated in whole or in part when the computer-executable instructions are loaded and executed on a computer. The computer instructions may be stored on a computer storage medium or transmitted from one computer storage medium to another via wireless (e.g., cellular, infrared, short-range wireless, microwave, etc.) to another website site, computer, server, or data center. The computer storage media may be any available media that can be accessed by a computer or a data storage device, such as a server, data center, etc., that incorporates one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., SSD), among others.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus described above may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A method of processing an image, comprising:
obtaining an image request, the image request comprising: a low-light image to be processed;
and according to the image processing request, adopting a pre-configured low-illumination enhancement network to brighten the low-illumination image to be processed, and acquiring a normal-illumination image corresponding to the low-illumination image to be processed.
2. The method of claim 1, further comprising:
acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
building a convolution network with a global illumination view;
and training the convolution network according to the training data set to obtain the low-illumination enhancement network.
3. The method of claim 2, wherein the obtaining a training data set comprises:
collecting a sample image of normal illumination;
respectively carrying out low-light processing on the collected normal-illumination sample images to obtain corresponding low-illumination sample images; and combining each of the normal-illumination sample images and the corresponding low-illumination sample image into a picture pair to obtain the training data set.
4. The method of claim 2, wherein said building a convolutional network with a global illumination field of view comprises:
performing global illumination distribution estimation processing on an input low-illumination image to obtain an up-sampling output result and a down-sampling output result;
combining the up-sampling output result and the down-sampling output result to obtain a combined result;
and sequentially carrying out convolution for three times and linear correction processing on the combined result to obtain an output normal illumination image and a corresponding brightening result.
5. The method of claim 4, wherein training the convolutional network to obtain the low-light enhancement network according to the training data set comprises:
and training the convolution network according to the training data set, the output normal light image and the corresponding brightening result thereof to obtain the low-light enhancement network.
6. An apparatus for processing an image, comprising:
an image request obtaining module, configured to obtain an image request, where the image request includes: a low-light image to be processed;
and the brightening processing module is used for brightening the low-illumination image to be processed by adopting a pre-configured low-illumination enhancement network according to the image processing request to obtain a normal-illumination image corresponding to the low-illumination image to be processed.
7. The apparatus of claim 6, further comprising:
the training data set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of sample images with normal illumination and corresponding sample images with low illumination;
the convolution network building module is used for building a convolution network with a global illumination view;
and the training module is used for training the convolution network according to the training data set so as to obtain the low-illumination enhancement network.
8. The apparatus of claim 7, wherein the training data set acquisition module comprises:
the sample image acquisition unit is used for acquiring a sample image under normal illumination;
the low light processing unit is used for respectively carrying out low light processing on the collected sample images with normal illumination to obtain corresponding sample images with low illumination; and combining each of the normal-illumination sample images and the corresponding low-illumination sample image into a picture pair to obtain the training data set.
9. An apparatus for processing an image, comprising: a memory, a processor;
a memory; a memory for storing the processor-executable instructions;
wherein the processor is configured to perform the method of any one of claims 1-5 by the processor.
10. A computer-readable storage medium having computer-executable instructions stored therein, which when executed by a processor, implement the method of processing an image according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810735358.7A CN110689486A (en) | 2018-07-06 | 2018-07-06 | Image processing method, device, equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810735358.7A CN110689486A (en) | 2018-07-06 | 2018-07-06 | Image processing method, device, equipment and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110689486A true CN110689486A (en) | 2020-01-14 |
Family
ID=69107290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810735358.7A Pending CN110689486A (en) | 2018-07-06 | 2018-07-06 | Image processing method, device, equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110689486A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931817A (en) * | 2020-07-10 | 2020-11-13 | 首钢集团有限公司 | Pellet ore phase identification method and device |
| CN112802137A (en) * | 2021-01-28 | 2021-05-14 | 四川大学 | Color constancy method based on convolution self-encoder |
| CN113256501A (en) * | 2020-02-10 | 2021-08-13 | 武汉Tcl集团工业研究院有限公司 | Image processing method, storage medium and terminal equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105046663A (en) * | 2015-07-10 | 2015-11-11 | 西南科技大学 | Human visual perception simulation-based self-adaptive low-illumination image enhancement method |
| CN107563984A (en) * | 2017-10-30 | 2018-01-09 | 清华大学深圳研究生院 | A kind of image enchancing method and computer-readable recording medium |
| CN108154547A (en) * | 2018-01-17 | 2018-06-12 | 百度在线网络技术（北京）有限公司 | Image generating method and device |
-
2018
- 2018-07-06 CN CN201810735358.7A patent/CN110689486A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105046663A (en) * | 2015-07-10 | 2015-11-11 | 西南科技大学 | Human visual perception simulation-based self-adaptive low-illumination image enhancement method |
| CN107563984A (en) * | 2017-10-30 | 2018-01-09 | 清华大学深圳研究生院 | A kind of image enchancing method and computer-readable recording medium |
| CN108154547A (en) * | 2018-01-17 | 2018-06-12 | 百度在线网络技术（北京）有限公司 | Image generating method and device |
Non-Patent Citations (3)
| Title |
|---|
| JIANRUI CAI ET AL.: "Learning a Deep Single Image Contrast Enhancer from Multi-Exposure Images", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| WENJING WANG ET AL.: "GLADNet: Low-Light Enhancement Network with Global Awareness", 《2018 13TH IEEE INTERNATIONAL CONFERENCE ON AUTOMATIC FACE & GESTURE RECOGNITION (FG 2018)》 * |
| 王书民 等: "一种基于全局和局部光照估计的Retinex图像增强算法", 《科学技术与工程》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113256501A (en) * | 2020-02-10 | 2021-08-13 | 武汉Tcl集团工业研究院有限公司 | Image processing method, storage medium and terminal equipment |
| CN111931817A (en) * | 2020-07-10 | 2020-11-13 | 首钢集团有限公司 | Pellet ore phase identification method and device |
| CN112802137A (en) * | 2021-01-28 | 2021-05-14 | 四川大学 | Color constancy method based on convolution self-encoder |
| CN112802137B (en) * | 2021-01-28 | 2022-06-21 | 四川大学 | Color constancy method based on convolution self-encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wei et al. | Deep retinex decomposition for low-light enhancement | |
| Wronski et al. | Handheld multi-frame super-resolution | |
| CN110163237B (en) | Model training and image processing method, device, medium and electronic equipment | |
| KR101699919B1 (en) | High dynamic range image creation apparatus of removaling ghost blur by using multi exposure fusion and method of the same | |
| WO2014169579A1 (en) | Color enhancement method and device | |
| JP2011028588A (en) | Information processing apparatus, line noise reduction processing method, and program | |
| CN110689486A (en) | Image processing method, device, equipment and computer storage medium | |
| Jeong et al. | Multi-frame example-based super-resolution using locally directional self-similarity | |
| JP5765893B2 (en) | Image processing apparatus, imaging apparatus, and image processing program | |
| Singh et al. | A novel approach for detail-enhanced exposure fusion using guided filter | |
| CN111242860A (en) | Super night scene image generation method and device, electronic equipment and storage medium | |
| Zamani et al. | Multiple-frames super-resolution for closed circuit television forensics | |
| CN111353982B (en) | Depth camera image sequence screening method and device | |
| CN111083359B (en) | Image processing method and apparatus, electronic device, and computer-readable storage medium | |
| KR101437898B1 (en) | Apparatus and method for generating a High Dynamic Range image using single image | |
| KR101341617B1 (en) | Apparatus and method for super-resolution based on error model of single image | |
| WO2014165159A1 (en) | System and method for blind image deconvolution | |
| JP2015179426A (en) | Information processing apparatus, parameter determination method, and program | |
| WO2023215371A1 (en) | System and method for perceptually optimized image denoising and restoration | |
| WO2016051716A1 (en) | Image processing method, image processing device, and recording medium for storing image processing program | |
| CN110992284A (en) | Image processing method, image processing apparatus, electronic device, and computer-readable storage medium | |
| JP6938282B2 (en) | Image processing equipment, image processing methods and programs | |
| Lin et al. | Single image super-resolution based on local self-similarity | |
| JP6584173B2 (en) | Image processing apparatus and image processing method | |
| CN116563190B (en) | Image processing method, device, computer equipment and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200114 |