Research On Algorithm Of H.264and3D Mesh Image Compression

H.264can achive very good coding performance now,but it applies high complexity and low complexity option,optional entropy coding method,which results in very high complexity.Deblocking also needs large buffer.These are not a problem for software platform.But they are problems for the hardware platform whose CPU frequency is low,whose buffer is limited.It needs very much work for optimization.This paper proposes optimization methods of H.264on hardware platform. At first this paper explores and studies innovation and applicationalgorithm in video compression standard H.264. Then, for the aspect of3D image compression, the author makes introduction and the innovation algorithm based on3D mesh compression. For the aspect of digital watermark, the author focuses on the study of the semi-fragile watermarking in H.264. For the aspect of subjective video quality evaluation and detection,this paper focuses on the detection of mosaic video images.The variance of U component based method is proposed.This paper begins with the method of Double Path Mode Decision.The innovation point is to use DSP platform under the condition of low complexity, inter macroblock mode decision will be assigned to two CPUs, which has greatly improved the speed of mode decision, and considers the close relationship between bitrate and DCT coefficients. The method reduces the bitrate by5%-25%under the same QP condition.In the session of3D image compression, although existing algorithm for topology, geometry, texture can achive some compression ratio,it needs much work on optimization.To explore the redundancy and optimization on3D image compression,this paper researches on entropy coding of topology, geometry, texture.This paper begins with new entropy coding method in the texture coding with maximum consumed complexity. A means of dynamic reference for floating-point numbers compression is proposed, floating-point numbers with high frequency are stored in a buffer, so if emerging data is same with some of the data in the buffer, then the buffer index is encoded. This algorithm is similar to the motion search in video, and finds the texture coding redundancy. Then the method of further compressing binary stream is proposed in the topology coding. According to the rule of code for topology coding,5symbols are reclassified.It puts C and R together to encode,and eliminates many combines of0and1. Re-encoding will reduce the entropy compared with original coding method, which can reduce the bitstream.Finally a new entropy coding method is proposed to encode a series of positive numbers in the geometry coding, firstly, it encodes whether the number is0. If it is not0, it encodes the number symbol, and then it encodes the absolute value.When the absolute value encoded, the number is expressed as binary at first. Then, mixed Columbus encoding method is applied. In this invention, sign of0, symbol signs, as well as binary mixed Columbus coding is encoded in different context models. In this invention, before context model based on self-adaptation binary arithmetic coding, XOR operation is applied.XOR operation means comparing the current bit and prior bit under the same context model.In the session of watermarking, H.264is the main video compression standard for network video transmission, so how to protect H.264bitstream to make it transmit from sender to receiver safely is a hot spot.This paper researches on the method of watermarking in H.264compression domain.This paper begins with1st method of Semi-fragile watermarking in compressed domain to protect DCT. Considering the protection of right signs of DCT, we adjust the number of positive and negative of DCT to watermarking dynamicly,considering the quality of image,we only change the symbol of DCT with the highest frequency to limit the impact on image quality.This method is simple and easy to use.Then2nd method of Semi-fragile watermarking in compressed domain to protect DCT is proposed, considering from the protection of right values and signs of DCT, we create a watermark matrix, the scope of each DCT coefficient changes is limited in two, and protect DCT coefficients of0and1, making it have minimized impact of the image. However, this method is very complex and requires updating watermark matrix constantly.It’s hard to take into practice on the platform with poor performance, but it is better in the aspect of watermark effect. The impact on image quality is the minimum while protecting the bitstream, and it is very practical on the platform with high performance.In the unit of subject video quality evaluation and detection, the main factors affecting the subject video quality are blocking, jitter, blur, noise. Mosaic as a special blocking effect damages the subject video quality very much.It’s important to detect the image and video with mosaic accurately.Based on the analysis of existing mosaic detection method,this passage proposes a new method based on the variance of U component.From the experiment result,it can be concluded that the new method has some advantages on error detection rate,leak detection rate and time consuming compared with existing spatial and edge domain method.It can achive normal commercial standard.At last, in the chapter five of the session for summarizing and prospecting, the author summarizes innovation points of this paper and describes the following jobs.