Research And Implementation Of Quality Assessment For Stereoscopic Images

The generation, compression, transmission and display technology of stereoscopic images has become focus at home and abroad. Compared with viewing 2-D images, a more immersive experience can be given by viewing stereo images. Yet, there has not been a complete stereo image quality assessment system to evaluate whether the quality of the stereo images is good or not. Therefore, it is significant to propose an accurate and efficient objective quality assessment method for stereo images, as a means or standard to measure the effect of stereo image processing.We deeply studied human visual characteristics and stereo imaging theory based on theoretical study and practical experiments, established a human visual model which reflected luminance sensitivity, contrast sensitivity, multi-channel effects and masking effects, and extracted six concrete evaluation metrics reflected luminance, signal to noise ratio and stereo sense. In order to verify the accuracy of the human visual model and whether the metrics could reflect the stereo image quality, we proposed a reasonable total evaluation metric by appropriate weighted method. The result showed that the object score calculated by our method had a high correlation with subjective score, and the correlation was up to 90%, which could reflect stereo image quality correctly.In order to make this stereo image quality assessment method easy and practical to use, we developed a software system, which could display image quality directly on the screen and also could save the metrics in a file for data analysis.