| With the rapid development of internet, the electronic consumable, communication, television, movie, broadcast and computer technology is connected more closely and the integrate of the communication and multi-media with the computer technology is an irreversible trend, which makes the multi-media industry based on internet become one of the industries that grow fastest and have the largest size. It is well known that the information human get by vision is about 70% of the total amount of information, and video information is intuitive and reliability, therefore, video technology has become one of the most important multi-media technologies. In order to store and transmit the digital video, compression is necessary for video. There are two characteristics for video signals to be compressed. Firstly, there is a large amount of redundancy and this redundancy can be resumed without distortion after encoding and decoding. Secondly, because of the vision characteristics of human, the gray-level of the quantified signal can be reduced in the condition that the images changes are not aware of, and get data compression in return by a certain objective distortion.With the widely application of video technology, a variety of compression standards were born. AVS group was found in June 2002, aiming at building up a China-owned IPR video codec standard, which could achieve higher encoding efficiency, and would be suited to applications with various bit rates such as in digital TV, video storage as well as video transmissions via networks. AVS has the similar technology framework as H.264, and performs as same as H.264, but is more concise in technology. AVS System use the program of MPEG-2 System, which is easy to be compatible with the MPEG-2 transport System.When dealing with the smooth images with no noise, AVS performances well and can achieve a good compression ratio. Therefore, the removal of noise in the video sequences is the most basic and the most important technology in the image analysis and computer vision and digital video processing. With the widely using of digital cameras, camcorders and video surveillance, the pre-processing of videos and images has become a very important technology. The design of the pre-processor should not only take the vision effects of encoded images into account, but also take the influence to the compression ratio of encoder. The ultimate goal is to make the vision effects of the video sequences become better after decoding, and make the encoded data transmit easily.The research on pre-processing algorithm of video based on AVS video coding standard in this paper can be divided into the following parts:1.This paper introduces some knowledge about video including the types of the color space, color TV sampling formats and the noises in the videos, which make the basic knowledge about video become more deeply. Secondly, several video pre-processing algorithms is studied including histogram balance algorithm, linear neighborhood average algorithm, nonlinear filtering algorithms, and low-pass filtering in frequency domain. We get to know the characteristics of the algorithms as well as their advantages and disadvantages.2.Study the technology of AVS video compression coding standard deeply. AVS use the framework as similar as H.264, including conversion, quantization, entropy coding, inter-frame prediction, intra-frame prediction, loop filter and so on.AVS standard also defines I frames, P frames and B frames. In the AVS video standard, all the blocks should be predicted inter-frame or intra-frame. Predicted residence should be 8×8 integer conversed and quantized, and quantitative coefficient should be scanned in the zig-zag form to get one dimensional array of quantitative coefficient,and be entropy encoded. The conversion and quantization only needs addition and subtraction and shifting, which can be completed with 16 precision. AVS uses loop filter on the reconstruction images, which can eliminate block affection, and improve the subjective quality of the images, as well as improve the coding efficiency.AVS video standard is based on the standards that were constructed by our innovative and international open technology. Its coding efficiency is as same as H.264, but its algorithm's complexity and cost is less than H.264. VI 3.This paper introduces a real-time video pre-processing filter algorithm using adaptive neighborhood statistics. This algorithm can be used in all kinds of real-time video application fields and related areas. It applied the concept of adaptive neighborhood statistics to video signal pre-processing. It makes use of the advantage of the linear average filter and nonlinear median filter and not only has good restraint characteristics for both additive Gauss and impulsive noise, but also retains the edge information of color images quite well. Besides, the low computing complexity of this algorithm makes it suitable for many kinds of real-time fields.In this paper, this algorithm is compared with scalar and vector median filter. The three filtering algorithms are all simulated, and the results of the simulation are observed in the condition when the two noise intensity are both large, and the impulsive noise intensity is larger than the Gauss noise, and the Gauss noise intensity is larger than the impulse, and the two are both small. The PSNR of the three are compared. In the subjective results, the algorithm of this paper gets better vision quality and maintains more details of the images. The two median filter both make images fuzzy, and have bad restraint characteristics for white Gauss noise. On the complexity, the algorithm of this paper is lower than the two median filter, and more suitable for real-time application.4. We analyse the performance of the algorithm of this paper on several points of view. Firstly, the vision effects and PSNR of the video sequences which is encoded by AVS video standard after decoding is compared with scalar and vector median filter. We can see that no matter when in the YUV color space or in the RGB color space, the algorithm of this paper improves the vision quality and objective PSNR, and reduces the noise. And then, we make 5 frames of the video sequence to be encoded by AVS standard, and compare the size of the video stream of the sequence that processed by the three algorithms. It shows that, the video sequence which is processed by the algorithm of this paper, whose size of the video stream is smaller noisy sequence. Its reduction is the most in the three algorithms. It shows that the pre-processing algorithm of this paper is more effective to remove the meaningless details in the video sequence, and the video stream is more propitious to transmit.5. This paper proposes an algorithm that we could adjust QP based on the times that an image is processed with neighborhood average algorithm when we use adaptive neighborhood algorithm. When the compression ratio is small, encoder prefers good vision quality of the images, but when it is large, encoder prefers a small video stream. This paper does a lot of simulation, and choose a threshold G. When the times of processing with neighborhood average algorithm is less than G, the detail of the image is kept, so we choose a smaller QP; when the time of processing with neighborhood average algorithm is more than G, the image is blurred, so we choose a larger QP. |
PDF Full Text Request