coderzie/imageproccessing

Image Processing Pipeline Architecture

Block Diagram

Module 1: Image Smoothening

• Functionality: This module processes 18 pixels in parallel.

• Timing Report:

  • DataPathDelay = 19.575
  • LogicDelay = 7.698
  • RoutingDelay = 11.877

• Utilization Report:

  • DynamicPower = 47.566
  • StaticPower = 1.31
  • SignalPower = 11.152
  • LogicPower = 11.878

• Power Report:

  • TotalPower = 48.876
  • Muxes = 26
  • LUT_as_memory = 124
  • LUT_as_logic = 390
  • Slice_LUT = 338
  • LUT1 = 76
  • LUT2 = 50
  • LUT3 = 0

• Simulation Results

Module 2: Image Sharpening

• Functionality: This module processes 9 pixels in parallel.
• Timing Report:
  • DataPathDelay = 10.803
  • LogicDelay = 7.590
  • RoutingDelay = 16.955
• Utilization Report:
  • DynamicPower = 24.798
  • StaticPower = 0.348
  • SignalPower = 6.626
  • LogicPower = 4.709
• Power Report:
  • TotalPower = 21.978
  • Muxes = 662
  • LUT_as_memory = 304
  • LUT_as_logic = 172
  • Slice_LUT = 18
  • LUT1 = 297
  • LUT2 = 0
• Simulation Results

Module 3: Image Addition

• Functionality: This module processes 1 pixel in parallel.
• Timing Report:
  • Slack: inf
  • DataPathDelay = 4.504ns
  • LogicLevels: 2
• Utilization Report:
  • Slice LUTs: 8
  • Registers: 11
• Power Report:
  • TotalPower = 0.047
• Simulation Results

Module 4: Normalization

• Functionality: This module processes 1 pixel in parallel.
• Timing Report:
  • DataPathDelay = 12.313
  • LogicDelay = 5.17
  • RoutingDelay = 7.143
• Utilization Report:
  • DynamicPower = 11.788
  • StaticPower = 0.185
  • SignalPower = 0.663
  • LogicPower = 0.564
• Power Report:
  • TotalPower = 11.973
  • Muxes = 104
  • LUT_as_memory = 7
  • LUT_as_logic = 43
  • Slice_LUT = 27
  • LUT1 = 10
  • LUT2 = 14
  • LUT3 = 16
• Simulation Results

Note: Detailed Synthesis Report are attached in the synthesis folder.

Design Pipeline

• Bit Width: 8 for BRAM
• Pipeline Functionality:
  • Initial Smoothening and Sharpening: Stall pipeline until 9 values are obtained.
  • Subsequent Processes: Stall pipeline until 2 values are read.
  • No stall required after the first two processes as only one value is needed.

Stages of the Pipeline

1. Stage 1: Smoothening - Part 1

   • Calculate sum of input pixels.

2. Stage 2: Smoothening - Part 2

   • Divide sum by total number of pixels (9 for 3x3 average filter).

3. Stage 3: Sharpening - Part 1

   • Perform convolution by multipling parallely.

4. Stage 4: Sharpening - Part 2

   • Perform accumulation of pixels.

5. Stage 5: Addition

   • Add sharpened image to input image.

6. Stage 6: Normalization - Part 1

   • Evaluate expression: 255×(Input_channel pixel value−min pixel value).

7. Stage 7: Normalization - Part 2

   • Divide product by (max pixel value−min pixel value).

Implementation Details

• Smoothening: Calculate two pixel values in parallel within a single clock cycle to support subsequent sharpening.
• Sharpening: Single clock cycle operation.
• Addition: Single clock cycle operation.
• Normalization: Implemented with two stages - numerator calculation and division, each taking one clock cycle.

This pipeline architecture optimizes processing efficiency while maintaining synchronization across modules.