Real-time Filters

Apply Instagram/TikTok-style filters to live video.

View Source Code

Overview

Apply various visual filters to webcam feed in real-time, similar to social media apps.

Key Techniques:

  • Custom convolution kernels
  • Color manipulation
  • Image blending
  • Look-up tables (LUTs)

Available Filters

Filter Technique Effect
Grayscale Color conversion Black and white
Sepia Matrix multiplication Vintage brown tone
Negative Inversion Inverted colors
Sketch Edge detection + threshold Pencil drawing
Emboss Custom kernel 3D relief effect
Sharpen Custom kernel Enhanced edges
Blur Gaussian filter Soft focus
Warm Channel adjustment Orange/yellow tint
Cool Channel adjustment Blue tint
Vignette Gradient mask Dark corners

Key OpenCV Functions 

# Grayscale
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

# Sepia (matrix multiplication)
sepia_kernel = np.array([
    [0.272, 0.534, 0.131],
    [0.349, 0.686, 0.168],
    [0.393, 0.769, 0.189]
])
sepia = cv2.transform(frame, sepia_kernel)

# Custom kernels
emboss_kernel = np.array([[-2, -1, 0], [-1, 1, 1], [0, 1, 2]])
embossed = cv2.filter2D(frame, -1, emboss_kernel)

# Sketch effect
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
inv = 255 - gray
blur = cv2.GaussianBlur(inv, (21, 21), 0)
sketch = cv2.divide(gray, 255 - blur, scale=256)

Vignette Effect 

def create_vignette(shape, strength=0.5):
    rows, cols = shape[:2]
    X = cv2.getGaussianKernel(cols, cols * strength)
    Y = cv2.getGaussianKernel(rows, rows * strength)
    mask = Y * X.T
    mask = mask / mask.max()
    return mask

Controls

Key Action
1-9, 0 Select filter
n Next filter
p Previous filter
s Save screenshot
q Quit

Running the Application 

python curriculum/applications/03_realtime_filters.py

Official Documentation