Research on 3D integer SDCT transform and quantisation

Abstract

Nowadays, the 3D discrete cosine transform (DCT) is applied widely in video coding. But the transform matrix of DCT is expressed with floating-point numbers, so the computational complexity is high, and more system resources are occupied. In addition, the 3D DCT is accomplished by operating 1D DCT to the rows, columns and pages of 3D data successively, which cannot embody the overall space performance of 3D transform well. To overcome these drawbacks, 3D integer submatrix discrete cosine transform (SDCT) method was proposed in the paper. First, several matrix operation methods were defined. Then, the basic principle and calculation method of 3D integer SDCT was deduced in detail. The main idea was to take the 3D data as a whole, and adopt the integer transform matrix instead of the floating-point transform matrix. Finally, the performances of 3D integer DCT operation were analysed, and the experimental results show that the transform effects based on 3D integer SDCT and 3D DCT are very similar.     

Multiplier Free IDCT Engine at MPEG-1 Video Data Rate

0~chopInspiteofcurmntbarkintheareaofstillimage.compressionandvideocompression['-'j,theJPEGandMPEGs}'stemsstinlankasthetWOdominantstandardsinstillimageandvideoimagecompression.FOrwaulandInverseDiscreteCosineTlansform.formanintegralpartofanyJPEGandMPEGsystem"-3).UPtonow,manydifferentalgoritLhlnstOdothediscretecosinet~formhavebeenproposed['-12].WhentakenintotheaccountofJPEGaccuracycompliance...id.rti..L13],thefastDCTalgorithmbyFeig[']reqUiresthedrinimumnumberofallthlneticoperahonsam…     

基于变换基阵的三维SDCT表示方法

对三维SDCT(sub-matrices discrete cosine transform)表示方法进行了研究。首先,定义了4种新的三维矩阵运算方法;其次,在构建了变换基阵的基础上,详细描述了三维SDCT的运算原理;最后,对三维SDCT算法特性进行了分析。通过理论分析表明,三维SDCT运算方法使三维DCT的表达简洁,理解容易,计算便捷。     

一种基于可重叠RNAM 的灰度图像表示算法

借助于二值图像的可重叠矩形区域编码的思想,通过使用可重叠矩形非对称逆布局的模式表示模型(RNAM)和扩展的Gouraud阴影法,给出了可重叠同类块逆布局的4个准则,提出了一种基于可重叠RNAM的灰度图像表示算法,简称为ORNAMC表示算法.在ORNAMC表示算法中,通过使用3个用于标识顶点类型的水平矩阵H、垂直矩阵V和单点矩阵I代替混合矩阵R,解决了灰度图像的可重叠RNAM表示中矩阵R的不可解码性问题;同时,通过将顶点类型及码字进行重新定义,提出了一种对矩阵H,V和I中所有非零元素坐标进行编码的坐标数据压缩算法.以图像处理领域里惯用的标准灰度图像等作为典型测试对象,实验结果表明,与已提出的非重叠RNAMC和流行的STC,SDCT等灰度图像表示方法相比,在保持图像质量的前提下,ORNAMC表示方法具有更高的压缩比和更少的块数,因而是灰度图像表示的一种更好的方法.