Universal no-reference image quality evaluation method based on multi-task convolutional neural network

Abstract

The invention discloses a no-reference image quality evaluation method based on a multi-task convolutional neural network, and belongs to the field of image perception. The method specifically comprises the following steps: step 1, extracting a plurality of image blocks with fixed sizes from each image of a manually labeled image quality data set, and each image block corresponding to two labels,which are respectively a degradation category of the image and a degradation degree of the image, so as to form a training set; step 2, constructing a convolutional neural network model based on dictionary learning; step 3, training the constructed convolutional neural network model by using the training set, and determining parameters of the convolutional neural network model after the training is finished; and step 4, during application, inputting the to-be-scored pollution image into the trained convolutional neural network model to obtain a corresponding image quality score. Compared witha traditional method, the method is higher in consistency with subjective evaluation in the field of non-reference image quality evaluation, and key indexes such as the Spearman rank correlation coefficient and the Pearson linear correlation coefficient are obviously improved.

Description

technical field [0001] The invention belongs to the fields of image perception and artificial intelligence, and in particular relates to a general non-reference image quality evaluation method based on a multi-task convolutional neural network. Background technique [0002] Images play an irreplaceable role in video communication, entertainment, and social networking. During the process of image acquisition, processing, transmission, and storage, images will inevitably be polluted by various noises, which will affect people's perception experience. Therefore, evaluating image quality is valuable for improving user experience. For example, a video provider can adjust image compression rate based on user feedback on image quality, so as to better utilize network transmission bandwidth, etc. Although human beings are the ultimate recipients of images, human subjective evaluation is the best option, but the process of subjective evaluation is time-consuming and labor-intensive, ...

AI Technical Summary

Problems solved by technology

However, it requires a large amount of subjectively evaluated image data for training

Images