Algorithm Research Of Color Image & H.26x Video Coding Based On D-APBT And Its Realization On ARM

Discrete cosine transform (DCT) plays an important role in major image and video compression nowadays. While at low bit rate, serious blocking artifacts exist while using DCT and the quantization is rather complicated. To solve these problems, all phase biorthogonal transform (APBT) was researched. All possible interception of phase is considered in APBT transform so that the blocking artifacts can be reduced effectively. In the process of APBT, the high frequency coefficients are attenuated, so the uniform quantization is used to quantify aspect, simplifying the quantization operation. For the images that horizontal and vertical edges not dominating, DCT may not be a good choice. Based on this, the directional DCT (D-DCT) which can take advantage of the directional information of image was researched. In order to improve the encoding performance of image and video at low bit rate and make full use of the directional information, color image coding and video coding algorithm based on directional APBT (D-APBT) is proposed in this thesis.Various direction modes and specific conversion of the D-APBT is firstly introduced in this thesis, and then it is used to the coding of color image and video. In the encoding process of color image, the color space conversion is firstly conducted. For each component converted, the coding part basically adopts JPEG framework. In the part of transform, D-APBT is used to replace the DCT. Instead of the complex table, uniform value which is also used to control the bit rate is adopted in the quantization step. Simulation results show that in the color image coding, D-APBT algorithm has obvious advantage. In the process of video encoding, D-APBT algorithm is respectively used in the H.263 and H.264 standards. In H.263, the proposed algorithm is used for the intra-coding, and then the quantization and entropy coding parts make some improvements. In H.264, the residual data of intra-coding adopts D-APBT algorithm. Then applying the improved algorithm to the coding of video sequences and simulation results show that D-APBT algorithm performs well at low bit rate.In order to verify feasibility of the proposed algorithm in reality, the improved H.264 algorithm is transplanted to the ARM board. To run the applications on the board, the migration of Linux kernel and root file system is firstly carried out. In the application level, the video capture program and video display program are combined with the improved video coding algorithm. Then the complete video codec application is accomplished. Running the application on the hardware environment, it could be found that the improved H.264 encoding algorithm achieves a good work by comparing the decoding frame with the original one.