Image quality evaluation method and apparatus, device and storage medium

1 . An image quality evaluation method, comprising:

acquiring an input image; and

inputting the input image into an image quality evaluation network to obtain an image quality evaluation result,

wherein the image quality evaluation network is configured to extract a first image feature from the input image, perform a shift operation on the first image feature to acquire one or more second image features, and determine the image quality evaluation result by combining the first image feature and the acquired second image features;

wherein a size of each of the second image features is the same as a size of the first image feature, and regions with identical values of features exist in different positions between the first image feature and the second image features;

wherein the image quality evaluation network comprises N cascaded predetermined modules, wherein N≥2; for an i th predetermined module, 1≤i≤N−1, and an output of the i th predetermined module is cascaded to an input of an (i+1) th predetermined module;

the predetermined modules are configured to perform a shift operation on an input image feature to acquire one or more third image features, and further extract image features by combining the input image feature and the acquired third image features and output the image features, wherein a size of each of the third image features is the same as a size of the input image feature, and regions with identical values of features exist in different positions between the input image feature and the third image features;

performing the shift operation on the first image feature to acquire the one or more second image features, and determining the image quality evaluation result by combining the first image feature and the acquired second image features comprises:

inputting the first image feature into a first predetermined module in the N cascaded predetermined modules to obtain an image feature output from an N th predetermined module; and

determining the image quality evaluation result according to the image feature output from the N th predetermined module.

2 . The method according to claim 1 , wherein determining the image quality evaluation result by combining the first image feature and the acquired second image features comprises:

determining one or more image attribute evaluation results by combining the first image feature and the acquired second image features; and

determining the image quality evaluation result according to the determined image attribute evaluation results.

3 . The method according to claim 1 , wherein performing the shift operation on the first image feature to acquire the one or more second image features comprises:

performing a shift operation on features in a predetermined region in the first image feature to obtain one or more shift results; and

filling a predetermined value for a default part in each of the one or more shift results to acquire the one or more second image features each having the same size as the first image feature.

4 . The method according to claim 1 , wherein determining the image quality evaluation result by combining the first image feature and the acquired second image features comprises:

further extracting space feature information of the input image by combining the first image feature and the acquired second image features, and

determining the image quality evaluation result according to the extracted space feature information.

5 . The method according to claim 1 , wherein the first image feature comprises at least one of: an image detail feature, an image noise feature, or an image global feature.

6 . The method according to claim 1 , wherein extracting the first image feature from the input image comprises:

acquiring predetermined information from the input image, and further extracting one or more image features from the predetermined information; and

adding the extracted image features to the first image feature;

wherein the predetermined information includes at least one of: an original image, image detail information, image noise information, image luminance information, image saturation information, or image hue information.

7 . The method according to claim 6 , wherein acquiring the predetermined information from the input image comprises:

performing edge filtering on the input image to extract the image detail information; and/or

performing guided filtering on the input image to obtain a denoised blurred image, and extracting the image noise information by combining the input image and the blurred image.

8 . The method according to claim 6 , wherein further extracting one or more image feature from the predetermined information comprises:

extracting one or more image features from different predetermined information by using different feature extraction networks.

9 . The method according to claim 1 , wherein a training set of the image quality evaluation network is generated by:

acquiring an unlabeled image sample;

for the unlabeled image sample, calculating one or more corresponding image attribute values; and

determining a corresponding image quality evaluation label according to the calculation result.

10 . An electronic device, comprising:

at least one processor; and

a memory connected in communication with the at least one processor,

wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to implement the method according to claim 1 .

11 . A non-transitory computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the method according to claim 1 .