Study On Objective Quality Assessment Method For Digital Video

Digital video quality assessment plays important roles in video compression, visual processing and video communications. In this dissertation, the objective digital video quality assessment is studied in depth, and several novel objective video quality assessment methods are proposed based on the characteristics of the Human Visual System (HVS). In addition, we also propose an error-concealment method for video reconstruction based on watermark technique, and an improved method which is suitable for gradient-threshold edge detectors.The major contributions of this dissertation are summarized as follows:1. The assessment of blocking artifacts of reconstructed video is studied firstly, and a simple method for measuring perceptive blocking artifacts taking account of the characteristics of luminance masking and activity masking is presented. Then, a new concept of blocking artifact cluster is proposed. Considering eye movement and fixation, some clusters with most serious blocking artifacts are pooled and used to assess the total blocking artifacts of the reconstructed video.2. Based on the analysis of blocking artifacts, a novel method for reducing blocking artifacts at the coding side is proposed. The quantization of the DCT coefficients is adjusted in the proposed method to smooth the quantization errors of spatially adjacent blocks during encoding.3. A novel objective no-reference metric is proposed for video quality assessment of digitally coded videos containing natural scenes. Taking account of the temporal dependency between adjacent images of the videos and the characteristics of the human visual system, the video quality is assessed using the differences between the corresponding translational regions of high spatial complexity in adjacent images.4. A reduced-reference video quality assessment method is proposed according to the characteristics of fragile watermark. The watermark information is embedded in the encoded video stream. The reconstructed video quality can be assessed by comparison between the watermark extracted from the transmitted video and the copy of the original watermark in the video decoder.5. Utilizing the capability of digital watermark to convey side information, an effective error concealment method for reconstructed video is proposed. The information of motion vectors is hidden in the coded video stream. At the decoder, error detection and error concealment are performed using the extracted information of motion vectors.6. A new full-reference video quality assessment method is proposed, where the correlation coefficient between the gradient of the reference image and the distorted image is used to assess the distortion of image content. The combination of the content distortion and the luminance distortion of the image is used to assess the image quality. The overall video quality is measured by pooling the quality of all images in the video.7. An improved method for the gradient-threshold edge detectors is presented. In this method the basic characteristics of HVS are taken into account and the gradient image is masked with the luminance and the activity of local image. Because the direction of gradient is easily affected by noise, it is very hard to correctly determine the regions for activity masking. An improved method for determining the regions for activity masking is proposed, by which the performance of gradient-threshold edge detectors is also significantly improved.