Image Quality Assessment Method Based On Structural Similarity With The MTF Study

Digital image is one of the most important ways for people to obtain information, and the image quality determines the accuracy and comprehensiveness to the information. Different types of distortion are introduced in the process of acquisition, transmission and processing for image, which decline the visual effect. As a result, how to measure image quality has become a basic problem in computer science and an important research topic.As subjective evaluation method affected by a variety of factors, it can not be applied to practical engineering. The traditional metric based on statistical theory is often unable to reflect the effect of human visual perception. Psychological research has shown that, there is different perception between image regions in human visual system. Therefore, it is great practical significance and application value to study on the metric based on visual perception features. The main work and contributions of this dissertation can be summarized as follows:(1) In order to analog multi-channel of HVS, the model uses two dimensional discrete wavelet transform of multi-scale to extract information, and then weights different image regions with Gaussian function and compares the differences between the original image and distorted image. Finally a multi-scale wavelet transform and weighted structural similarity evaluation metric (MSWSSIM) is proposed in this dissertation. The approach predicts the quality of various distorted images, and improves the performance when compared with traditional methods and has a better consistency with subjective evaluation value.(2) Human eyes have high sensitivity to image edge and texture. In this dissertation, a metric based on multi-scale wavelet transform and gradient information (MGSSIM) is proposed. Wavelet transform modulus maxima is used to detect image features, and then image regions are weighted with Gaussian function. This approach combines the human visual features, and is able to measure the image quality accurately and effectively. The experiment results show that the performance of this model has been improved greatly and its consistency with subjective evaluation value is better than other metrics.(3) After studying the common methods of computing the MTF of optical imaging system, edge sampling phase has been improved and applied in blade edge method. Along with MTF evaluation metric, the MSWSSIM, GSSIM and MGSSIM measurement model are tested on various kinds of remote sensing images. The experiment results show that the four models are able to analysis image quality efficiently from several different perspectives.