Image Quality Assessment And Classification Based On Artifacts

Objective image quality assessment is a classical topic. The goal is to assess the quality of image in agreement with human visual system. According to whether the reference image is available, it can be classified into three groups: Full-reference, reduced-reference and no-reference. No-reference image quality assessment has bright future for application. Present day no-reference image quality assessment usually assume that the distortion affecting the image is known. This is a limiting assumption for practical applications, since in a majority of cases the distortions in the image are unkown. This thesis presents a no-reference image quality metric, which doesn't require any knowledge of the distortion process. It is based on individual measurements of four artifacts affecting image quality: blockiness, blurriness, noisiness and ringring. Models for the overall annoyance based on a combination of the artifact metrics using Minkowski metric. The weights are trained from the subjective database.In addition,we have designed a method to classify the image based on artifacts. The artifacts occurred in distorted images under different types of process are different. Given a distorted image,we first measure the perceived values of these artifacts and then pool them into a feature vector ,which can be used to classify the image into different categories by SVM. Although we only consider four kinds of image process here, but it can be extended to any number of distortions by analysis the possible artifacts. Both methods are based on machine learning.