Skip to content

Filter‐Based Testing

Lucky8boy edited this page Aug 25, 2025 · 1 revision

🧪 Panel 5 - Filter-Based Testing

Panel 5 of SubSpectra focuses on visual and pixel-level analysis of images using a wide variety of filters. While not statistical in nature like Chi-square or RS, these effects help visually identify inconsistencies, compression artifacts, or steganographic alterations in LSB layers.


🎨 Filters and Transformations

All filtered images are saved to a folder named:

[original_image_name]_filters

in the same directory as the source image.

✅ Available Filters:

1. Channel Bit Removal (R / G / B / A)

  • Removes a specific bit layer (0–7) from the chosen channel (Red, Green, Blue, Alpha)
  • Example filename: image_R_bit_0_removed.png

2. Color Filters

  • Red, Green, Blue: isolates one channel by zeroing the others
  • Gray: average of R, G, B to produce grayscale
  • BW: binary black-and-white using a brightness threshold
  • Sepia: warm-toned retro effect (based on weighted average formula)

3. Block Averaging Filters

  • 3x3 and 9x9 block average filters replace each pixel with the average color of its neighborhood block
  • Highlights uniform or noisy areas

4. Special 3x3 Deviation Filter

  • For every pixel:

    • Computes average of surrounding 3x3 block
    • If pixel's R / G / B > average → set to 255
    • Else → leave unchanged
  • Emphasizes local anomalies that may suggest manipulation

  • Saved as: image_special_3x3.png


👁️‍🗨️ Use Case: Visual Forensics

  • Filters help reveal hidden structure, such as uniform LSB layers
  • Block filters expose macroblock compression or spatial inconsistencies
  • BW / Sepia filters emphasize edge changes and unnatural transitions

cat


📁 Output Folder Structure

Image.png
├── Image_filters/
│   ├── Image_R_bit_0_removed.png
│   ├── Image_G_bit_0_removed.png
│   ├── Image_B_bit_0_removed.png
│   ├── Image_A_bit_0_removed.png
│   ├── Image_R_bit_1_removed.png
│   ├── Image_G_bit_1_removed.png
│   ├── Image_B_bit_1_removed.png
│   ├── Image_A_bit_1_removed.png
......
│   ├── Image_R_bit_7_removed.png
│   ├── Image_G_bit_7_removed.png
│   ├── Image_B_bit_7_removed.png
│   ├── Image_A_bit_7_removed.png
│   ├── Image_red.png
│   ├── Image_gray.png
......
│   ├── Image_3x3.png
│   ├── Image_9x9.png
│   └── Image_special_3x3.png

🧩 Tips

  • Best used in combination with bit layer visualizer from Panel 3
  • Try removing suspected LSB layers and compare results
  • Run filters after decoding stego image to identify tampering