Studies On The Compressed Domain Image/Video Spatial Resolution Conversion And Color Processing Technology

With the advance of image and video compression techniques, most of the image and video streams are existed in the form of compressed format, so it is more important to develope the compressed domain processing technology. For example, in order to meet with the requirements of net bandwidth and display terminal device, it is necessary to convert the image spatial resolution in the compressed domain. Compressed domain image and video editing could provide the extensive supports for the enhancement of digital TV service and mobile phone digital multimedia service functionality expansion. This paper makes research on the key techniques in the compressed domain image and video processing technology, including compressed domain arbitrary spatial resolution conversion, compressed domain image luminance adjustment, color mixture rendering and color correction, compressed domain color saturation adjustment technologies. The thesis contributes the following new achievements:As regard to image spatial resolution conversion, we propose a new compressed domain arbitrary image resolution conversion method based on the relationship of the DCT coefficients between a block and its sub-blocks with differing size. It could realize the arbitrary unequal ratio image resolution conversion in the horizontal and vertical directions by the block recomposion and decomposition. It recomposed the image into the new one composed of the supporting areas. The resolution conversion is realized by the supporting area downsampling or upsampling. The experimental results indicate that our algorithm is low computational and memory efficient compared to the existing methods, it could also get better visual quality and high PNSR.As for the DCT domain image and video color adjustments editing, by analysising the relationship between the RGB and YUV color space in the DCT domain, we propose a linear luminance adjustment method, color mixture rendering method and color correction method respectively. These methods are all realized in the DCT domain of the YUV color space.The above three linear color processing algotithms realize the conventional color processing technologies in the RGB color space directly in the DCT domain of the YUV color space. They do not need full decompression, IDCT and the conversion of color space from YUV to RGB. It saves the computational complexity and memory to large extent. The experimental results show that during high quality image editing, the images and videos processed by the proposed method produce identical results visually compared to the corresponding method that realized in the spatial domain of RGB color space. The three methods are all processed independently on the DCT block basis and they are easier for hardware implementation.As for the DCT domain image color saturation adjustment, by analyzing the spatial domain luminance based adaptive color saturation adjustment, we propose a compressed domain color saturation adjustment method. Based on the mean value of each luminance block, it calculates the maximum chrominance vectors corresponding to the luminance block, so that the upper bound of allowed saturation factor for each chrominance vectors can be derived. The method is a block adaptive one. We use the block classification and segmentation to further improve the color saturation adjustment performance. It realizes the color saturation adjustment of blocks with differing characteristics by using blocks with differing size. The experimental results show that during high quality image editing, the images processed by the proposed method produce identical results visually compared to the spatial domain luminance based adaptive color saturation adjustment. The method is processed independently on the DCT block basis and it is computational and memory efficient and easier for hardware implementation.Most of the work in this thesis is supported by the project"HDTV Editing Card"established by the National Plan Committee.