Quality evaluation method without reference image based on regional mutual information

Abstract

The invention relates to the quality evaluation method without a reference image based on regional mutual information and belongs to the computer image analysis field. The method comprises the following steps: 1 carrying out 256*256 image blocking on an image so that image analysis is accurate; 2 acquiring regional mutual information RMI of an image block in each image block and dividing into 12 groups of RMI through distinguishing a size of a window, a motion direction and a distance size; 3 for each group of the RMI, describing a characteristic, carrying out wavelet decomposition processing on each group of the RMI, calculating a mean value of decomposed RMI low frequency information and calculating a variance of high frequency information; 4 using each dimension of RMI description characteristic to calculating a mean value of all the image block characteristics and taking the mean value as a characteristic of a whole image; 5 using a support vector machine and a method of combination of two step framework in image quality evaluation to calculating image quality. The method has advantages that subjective consistency is high; versatility is good and so on. The method can be used in an image processing correlation system and image equipment and possesses a good utility value.

Description

technical field [0001] The invention relates to a non-reference image quality evaluation method based on regional mutual information, which belongs to the field of computer image analysis. Background technique [0002] Human vision is the most basic and effective way of human perception, and images are a tool to record what people see visually at a certain moment. The information contained in the image allows people to directly and clearly obtain the recorded visual information, which is an advantage that other methods do not have. The related field of computer image analysis takes visual information as the research object, and studies how images can be better processed, so as to assist human related work. Therefore, computer image analysis research is a very important and active applied research field. The image records the visual state of a certain environment at a certain moment, but when the computer processes the image information, noise and errors often occur, such a...

AI Technical Summary

Problems solved by technology

[0013] Generally speaking, the current no-reference image quality evaluation methods still generally have the defects of low subjective consistency, high space complexity and long algorithm calculation time.

Images