A Tone Mapping Image Quality Evaluation Method Based on Structural Similarity Difference

Abstract

The invention discloses a tone mapping image quality evaluation method based on structural similarity difference. The method first extracts the mean value and variance of the gradient binarization matrix of the image as local gradient features, and then extracts the mean value, variance, and Kurtosis and skewness are used as local structural features, then the uniform local binary pattern histogram of the phase image is used as the global phase feature, and finally the local structural similarity (SSIM) difference between the neighboring pixel blocks and the central pixel block is used. The mean, variance, kurtosis, and skewness are used as the characteristics of the similarity difference of the local neighborhood structure, and the overall image quality evaluation feature is obtained by fusion, which is sent to the support training regression machine for training and testing, and the objective image quality evaluation result is obtained; this method avoids The disadvantage of low evaluation accuracy caused by using a single global feature or local feature is eliminated, and the image quality evaluation accuracy is improved.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a tone mapping image quality evaluation method based on structural similarity difference. Background technique [0002] Image quality evaluation is a key issue in the field of image processing. Image quality evaluation methods can be divided into subjective image quality evaluation methods and objective image quality evaluation methods according to whether people participate. Subjective image quality evaluation methods are scored by humans, and the evaluation results are accurate, but the evaluation process is complex, time-consuming, and difficult to be applied in real time. The objective image quality evaluation method does not require human participation, and the image quality is automatically predicted by a specific computer algorithm. According to whether the original undistorted image is used as a reference, the image quality evaluation method can be divided into a full refe...

AI Technical Summary

Problems solved by technology

[0003] High dynamic range (HDR) images display a large brightness range, which can bring better visual experience to viewers, but most of the existing graphics display devices only support displaying 8-bit low dynamic range (LDR) images, so it is necessary to use The tone-mapping algorithm converts the HDR image into an LDR image, and the LDR image converted from the HDR image becomes a tone-mapped image. During the tone-mapping process, the tone-mapped image has distortions such as contrast, naturalness, and color saturation, while block effects, blurring, etc. , white noise and other distortions are less, and the traditional no-reference image quality evaluation method is mainly effective for natural images and distortions such as block effects, blurring, and white noise, but the effect is not good for tone-mapped images. Reference image quality assessment is more challenging

No-reference quality assessment for tone-mapped images, Guanghui Yue [Guanghui Yue, Chunping Hou, and Tianwei Zhou, Blind Quality Assessment of Tone-Mapped Images Considering Colorfulness, Naturalness and Structure, IEEE Transactions ON Industrial Electronics, 2018.] extracts chroma-mapped distorted images Image quality assessment based on color, naturalness and structural features; Gangyi Jiang [Blind tone-mapped has been with the School of Electronic and image quality assessment based on brightest / darkest regions, naturalness and aesthetics, IEEE Access, vol.6, pp .2231-2240, 2018.] Use image natural statistical features and aesthetic features for quality evaluation; however, these methods focus on the natural statistical and global features of images, and most of them use traditional feature extraction methods that are effective for natural images, so the effect is similar to that of There are gaps in the subjective evaluation results

Images