JPEG Encoder

Hardware Implementation of JPEG Encoder for Real-time Compression

The software implementation aims to realize the JPEG compression algorithm from the lowest level

• it can be used to generate test vectors for verifying the hardware implementation on FPGA.
• it only requires matplotlib or opencv-python module to read the .bmp file and numpy for matrix operations.

The ASIC Implementation is fully synthesizable which contains non-blocking floating-point modules designed by HDL.

• floating-point multiplier (radix-4 booth algorithm, round to the nearest even)
• floating-point adder (round to the nearest even)
• floating-point divider (long-term division, round to zero)
• floating point to integer
• integer to floating point

Python vs FPGA Implementation

# Result

Row Image	Encoded by Python IMP.	Encoded by FPGA IMP.

Python Implementation

Print detailed coding process for each minimum coded unit.

DCT Matrix

[[ 0.354  0.354  0.354  0.354  0.354  0.354  0.354  0.354]
 [ 0.49   0.416  0.278  0.098 -0.098 -0.278 -0.416 -0.49 ]
 [ 0.462  0.191 -0.191 -0.462 -0.462 -0.191  0.191  0.462]
 [ 0.416 -0.098 -0.49  -0.278  0.278  0.49   0.098 -0.416]
 [ 0.354 -0.354 -0.354  0.354  0.354 -0.354 -0.354  0.354]
 [ 0.278 -0.49   0.098  0.416 -0.416 -0.098  0.49  -0.278]
 [ 0.191 -0.462  0.462 -0.191 -0.191  0.462 -0.462  0.191]
 [ 0.098 -0.278  0.416 -0.49   0.49  -0.416  0.278 -0.098]]

MCU: (0, 0)

 [[ 48 247 255 238 238 254 255 255]
  [ 96 239 238 238 206 238 238 238]
  [ 96 238 238 255 187 251 239 255]
  [ 96 255 238 255 206 238 190 187]
  [ 96 255 255 255 255 255 255 255]
  [ 96 238 238 238 222 255 255 255]
  [ 96 207 253 255 191 255 255 239]
  [ 64 236 254 175 136 136 216 253]]

Level shift:

 [[-80 119 127 110 110 126 127 127]
  [-32 111 110 110  78 110 110 110]
  [-32 110 110 127  59 123 111 127]
  [-32 127 110 127  78 110  62  59]
  [-32 127 127 127 127 127 127 127]
  [-32 110 110 110  94 127 127 127]
  [-32  79 125 127  63 127 127 111]
  [-64 108 126  47   8   8  88 125]]

FDCT:

 [[ 706.5   -197.001 -153.304 -225.402 -186.75   -82.355  -57.186  -77.932]
  [  41.361  -10.314  -42.856   28.413   -1.487  -19.936   -6.573  -21.584]
  [ -51.716  -25.931   40.813  -51.423  -19.048    2.665   -1.177   28.599]
  [  75.527  -49.779  -39.653   32.536   -1.707  -13.256   -6.264   13.657]
  [ -29.      25.074   -8.076  -28.362   -8.75   -37.158  -18.844   22.717]
  [   1.836    3.594  -36.801   17.595   -6.772   17.278   15.212  -25.5  ]
  [  -7.262  -17.817   25.573  -21.021   -4.828  -15.732   -6.563    2.629]
  [  44.865  -37.233   -2.265   -6.343    1.419   10.662    4.243    2.5  ]]

Quantized:

 [[ 44.156 -17.909 -15.33  -14.088  -7.781  -2.059  -1.121  -1.278]
  [  3.447  -0.86   -3.061   1.495  -0.057  -0.344  -0.11   -0.392]
  [ -3.694  -1.995   2.551  -2.143  -0.476   0.047  -0.017   0.511]
  [  5.395  -2.928  -1.802   1.122  -0.033  -0.152  -0.078   0.22 ]
  [ -1.611   1.14   -0.218  -0.506  -0.129  -0.341  -0.183   0.295]
  [  0.077   0.103  -0.669   0.275  -0.084   0.166   0.135  -0.277]
  [ -0.148  -0.278   0.328  -0.242  -0.047  -0.13   -0.055   0.026]
  [  0.623  -0.405  -0.024  -0.065   0.013   0.107   0.041   0.025]]

Round:

 [[ 44. -18. -15. -14.  -8.  -2.  -1.  -1.]
  [  3.  -1.  -3.   1.  -0.  -0.  -0.  -0.]
  [ -4.  -2.   3.  -2.  -0.   0.  -0.   1.]
  [  5.  -3.  -2.   1.  -0.  -0.  -0.   0.]
  [ -2.   1.  -0.  -1.  -0.  -0.  -0.   0.]
  [  0.   0.  -1.   0.  -0.   0.   0.  -0.]
  [ -0.  -0.   0.  -0.  -0.  -0.  -0.   0.]
  [  1.  -0.  -0.  -0.   0.   0.   0.   0.]]

Dequantized (Lossy):

 Check with the JPEGsnoop DCT Matrix
 [[ 704. -198. -150. -224. -192.  -80.  -51.  -61.]
  [  36.  -12.  -42.   19.   -0.   -0.   -0.   -0.]
  [ -56.  -26.   48.  -48.   -0.    0.   -0.   56.]
  [  70.  -51.  -44.   29.   -0.   -0.   -0.    0.]
  [ -36.   22.   -0.  -56.   -0.   -0.   -0.    0.]
  [   0.    0.  -55.    0.   -0.    0.    0.   -0.]
  [  -0.   -0.    0.   -0.   -0.   -0.   -0.    0.]
  [  72.   -0.   -0.   -0.    0.    0.    0.    0.]]

Zigzag:

 [[ 44 -18   3  -4  -1 -15 -14  -3]
  [ -2   5  -2  -3   3   1  -8  -2]
  [  0  -2  -2   1   0   0   0   0]
  [  1   0   0  -1  -1   0   0   0]
  [ -1  -1   0   1   0   0   0   0]
  [  0   0   0   1   0   0   0   0]
  [  0   0   0   0   0   0   0   0]
  [  0   0   0   0   0   0   0   0]]

RLE:

 - DC: 44
 - AC: [(0, -18), (0, 3), (0, -4), (0, -1), (0, -15), (0, -14), (0, -3), (0, -2), (0, 5), (0, -2), (0, -3), (0, 3), (0, 1), (0, -8), (0, -2), (1, -2), (0, -2), (0, 1), (4, 1), (2, -1), (0, -1), (3, -1), (0, -1), (1, 1), (7, 1), (0, 0)]

Huffman encoding:

 - DC: ((4, 14), (6, 44))
 - AC: [((5, 26), (5, 13)), ((2, 1), (2, 3)), ((3, 4), (3, 3)), ((2, 0), (1, 0)), ((4, 11), (4, 0)), ((4, 11), (4, 1)), ((2, 1), (2, 0)), ((2, 1), (2, 1)), ((3, 4), (3, 5)), ((2, 1), (2, 1)), ((2, 1), (2, 0)), ((2, 1), (2, 3)), ((2, 0), (1, 1)), ((4, 11), (4, 7)), ((2, 1), (2, 1)), ((5, 27), (2, 1)), ((2, 1), (2, 1)), ((2, 0), (1, 1)), ((6, 59), (1, 1)), ((5, 28), (1, 0)), ((2, 0), (1, 0)), ((6, 58), (1, 0)), ((2, 0), (1, 0)), ((4, 12), (1, 1)), ((8, 250), (1, 1)), ((4, 10), (0, 0))]
 - DC: ['1110-101100']
 - AC: ['11010-01101', '01-11', '100-011', '00-0', '1011-0000', '1011-0001', '01-00', '01-01', '100-101', '01-01', '01-00', '01-11', '00-1', '1011-0111', '01-01', '11011-01', '01-01', '00-1', '111011-1', '11100-0', '00-0', '111010-0', '00-0', '1100-1', '11111010-1', '1010-']

Byte stream:

 EB 34 D7 8C 58 58 A2 CA A8 E6 DD 76 A9 EF C0 E8 33 F5 

FPGA Implementation

The designed FPGA IP is fully synthesizable and has been implemented onto the Digilent Atry-Z7 development board (Xilinx Zynq xc7z020clg400-1) for demostration.

Post-Implementation Utilization

FPGA Implementation using floating-point IPs

FPGA Implementation using fully synthesizable IPs

Post-Implementation P&R

FPGA Implementation using floating-point IPs

FPGA Implementation using fully synthesizable IPs

Test Platform

Use external UART cable to challenge the designed JPEG encoder at baudrate of 115200.

!!! Reset is required before executing the python script jpeg_z7_streamer !!!

Result

Performance	CR	PSNR
Min	3.656	24.245
Max	5.821	27.652
Average	4.543	25.733