Chapter 18Image Compression and Hardware Extensions TI C5000官方课件.pptVIP

  • 2
  • 0
  • 约8.97千字
  • 约 22页
  • 2018-01-25 发布于浙江
  • 举报

Chapter 18Image Compression and Hardware Extensions TI C5000官方课件.ppt

Chapter 18Image Compression and Hardware Extensions TI C5000官方课件

Copyright ? 2003 Texas Instruments. All rights reserved. DSP C5000 Chapter 18 Image Compression and Hardware Extensions Image Compression Discrete Cosine Transform (DCT) based compression Still image: JPEG. Remove spatial redundancy Video: H263, MPEG4. Remove spatial redundancy Remove temporal redundancy JPEG Encoder Image is split into 8x8 pixel block that are processed . DCT is applied on input block to obtain the frequency data. Values are read from low to high frequency thanks to zig-zag addressing. Data are then quantized according to a quality factor. This output is then run length and entropy coded. JPEG Decoder Data are entropy decoded. They are dequantized according to a quality factor. Zig-zag addressing to re-order the data pixel by pixel. Inverse DCT to recover spatial data. H263 and MPEG4 encoder Same basic building blocks H263 and MPEG4 Decoder Same basic building blocks Hardware Extensions ‘C5509 and ‘C5510 provide H/W extensions for video compression: DCT, IDCT Pixel interpolation Motion estimation Hardware extensions are embedded co-processors: Access to ACx, DARAM and SARAM. Provide video codec with less than half of the processing bandwidth. DCT H/W Extension 1 of 5 Allows 4x4 and 8x8 pixels DCT/IDCT computation: 2D DCT 2D IDCT DCT H/W Extension 2 of 5 Basic steps to complete a 2D DCT/IDCT: Input Data: macro block (4x4 or 8x8) is read in (DCT). DCT coefficient matrix is read in (IDCT). Column DCT/IDCT process: output to temporary registers that will be transposed for Row processing. Row DCT/IDCT process: data read from temporary register which are processed column by column. DCT H/W Extension 3 of 5 DCT H/W Extension 4 of 5 Instruction set: Load+computation+transfer to accumulator: ACy=copr(k8,ACx,Xmem,Ymem) Computation+transfer to accumulator+memory write: ACy=copr(k8,ACx,ACy),Lmem=Acz Special instructions: ACy=copr(k8,ACx,ACy) Refer to application note for details and listings of macros realizing complete 8x8 DCT/IDCT. DCT H/W Exten

您可能关注的文档

文档评论(0)

1亿VIP精品文档

相关文档