Scribd
Upload
4 views2 pages

Image Processing Viva Notes

The document outlines various image processing techniques across multiple labs, including basic transformations, filtering methods, frequency domain analysis, noise models, edge detection, and image compression. Key methods discussed include negative image transformation, Gaussian filtering, Canny edge detection, and the Discrete Cosine Transform (DCT) for compression. Each lab focuses on specific techniques and their applications in enhancing and analyzing images.

Uploaded by

shivarajm.btech22
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views2 pages

Image Processing Viva Notes

The document outlines various image processing techniques across multiple labs, including basic transformations, filtering methods, frequency domain analysis, noise models, edge detection, and image compression. Key methods discussed include negative image transformation, Gaussian filtering, Canny edge detection, and the Discrete Cosine Transform (DCT) for compression. Each lab focuses on specific techniques and their applications in enhancing and analyzing images.

Uploaded by

shivarajm.btech22
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Lab 1: Basic Image Transformations

1. Negative Image: Inverts pixel intensities (s = L-1-r). Highlights bright regions.


2. Log Transformation: s = c*log(1+r); enhances dark pixels (c=1,5,perfect c).
3. Gamma Transformation: s = c*r^γ; γ<1 brightens, γ>1 darkens.
4. Contrast Stretching: Expands intensity range to enhance contrast.
5. Thresholding: Converts image to binary (0 or 255) using threshold T.
6. Alpha Blending: Combines two images: output = α*I1 + (1-α)*I2.

Lab 2: Filtering Techniques

1. Box Filter: Averages neighborhood pixels equally (blurs edges).


2. Gaussian Filter: Weighted average; smoother edges, less noise.
3. Min Filter: Removes white (salt) noise.
4. Max Filter: Removes black (pepper) noise.
5. Median Filter: Removes both salt and pepper; preserves edges best.

Lab 3: DFT and Convolution Theorem

1. DFT on patterns: Converts image to frequency domain; reveals frequency components.


2. Matrix DFT: W*A*W gives same result as [Link].fft2(A).
3. Convolution Theorem: Convolution in spatial = multiplication in frequency.

Lab 4: Frequency Domain Filtering

1. Ideal LPF: Sharp cutoff; strong blur; ringing occurs.


2. Gaussian LPF: Smooth cutoff; natural blur.
3. Butterworth LPF: Adjustable sharpness via order n.
4. HPF filters: Opposite of LPF; emphasize edges and details.

Lab 5: Noise Models and Restoration

Noise Models:
• Gaussian – random, bell-shaped.
• Impulse – salt & pepper noise.
• Uniform – equal probability.
• Rayleigh – positive skewed.
• Exponential – one-sided decay.

Filters:
• Arithmetic Mean – smooths Gaussian noise.
• Geometric Mean – preserves edges.
• Median (Order-statistics) – removes impulse noise.

Frequency Filters:
• Bandpass/Reject – pass or remove selected frequency bands (Ideal, Gaussian, Butterworth).

Wiener Filter – adaptive filter minimizing mean square error; balances sharpening and noise
reduction.

Lab 6: Edge Detection

1. Roberts Operator – 2x2 filter; simple; detects sharp edges; noise sensitive.
2. Prewitt Operator – 3x3 filter; detects vertical & horizontal edges.
3. Sobel Operator – weighted 3x3 filter; robust to noise; sharper edges.
4. Derivative of Gaussian (DoG) – subtracts blurred images; smooth and accurate edges.
5. Laplacian of Gaussian (LoG) – second derivative; detects fine detailed edges.

Lab 7: Canny Edge and Region Segmentation

Canny Edge Detector:


• Steps: Gaussian blur → Gradient (Sobel) → Non-Max Suppression → Double Threshold → Edge
Tracking.
• Produces thin, continuous edges; very accurate.

Region Splitting & Merging:


• Splitting: Divide non-homogeneous region into 4 parts until uniform.
• Merging: Combine adjacent homogeneous regions.
• Result: uniform segmented image.

Lab 8: Image Compression using DCT & IDCT

1. DCT converts spatial domain to frequency domain (energy compaction).


2. Most energy is in low-frequency coefficients.
3. Compression: zero out high-frequency components in 8x8 blocks.
4. IDCT reconstructs the image from compressed DCT data.
5. Compression ratio = (1 - compressed/original) × 100.
6. Basis for JPEG compression.

You might also like