Question 1

Q.1. (a) Discuss down sampling (or subsampling) and up sampling of an image w…

Accepted Answer

Down sampling reduces image resolution by discarding pixels (needing an anti-alias low-pass filter first), while up sampling increases resolution by inserting interpolated pixel estimates; a CCD sensor converts incident light into charge via the photoelectric effect and shifts it out through vertical and horizontal shift registers to an output amplifier and ADC to form the digital image.

Question 2

Q.1. (a) What are the basic components of a digital image processing system? …

Accepted Answer

A digital image processing system consists of image sensors, processing hardware, storage, display, and hardcopy/networking units and is applied in medical imaging, remote sensing, and industrial machine vision; sampling discretizes spatial coordinates while quantization discretizes intensity values, together converting a continuous image into a digital one.

Question 3

Q.2. Describe the various types of frequency domain filters. [16]

Accepted Answer

Frequency domain filters modify the Fourier transform of an image to achieve smoothing or sharpening; ideal filters give a sharp cutoff but cause ringing, Butterworth filters offer a smooth order-controlled transition with less ringing, and Gaussian filters give the smoothest transition with no ringing at all, with high-pass versions and band-pass/notch filters obtained analogously.

Question 4

Q.2. (a) Discuss image sharpening by using second order derivative with an ex…

Accepted Answer

Second-order derivative (Laplacian) sharpening highlights fine detail and edges by adding back the Laplacian response to the original image, while discrete histogram equalization cannot produce a perfectly flat histogram because the transformation is a discrete step-function approximation of the ideal continuous, uniform-output mapping.

Question 5

Q.3. (a) Discuss working of order static filter and their suitable applicatio…

Accepted Answer

Order-statistic filters (median, max, min, midpoint) rank the pixel values within a neighborhood and select a specific ranked value for noise removal, and the adaptive median filter extends this by dynamically adjusting its window size based on local statistics to remove higher noise densities while preserving detail better than a fixed-size median filter.

Question 6

Q.3. (a) Explain the noise estimation parameter. [8]

Accepted Answer

Noise is estimated from flat image regions using mean, variance, SNR, and histogram/Fourier analysis; inverse filtering recovers the original image by dividing by the degradation function but amplifies noise badly, whereas Wiener filtering balances restoration and noise suppression using a noise-to-signal power ratio term.

Question 7

Q.4. (a) Discuss image opening morphological operation by writing its mathema…

Accepted Answer

Opening (erosion followed by dilation, A o B = (A minus B) plus B) removes small protrusions and thin connections while preserving overall object shape; closing (dilation followed by erosion, A bullet B = (A plus B) minus B) fills small holes and gaps while similarly preserving overall shape, each smoothing contours from opposite sides.

Question 8

Q.4. (a) Explain Erosion and Dilation. [8]

Accepted Answer

Erosion shrinks object boundaries (removing small objects and thin protrusions) while dilation grows object boundaries (filling small holes and gaps), each formally defined via set-translation of a structuring element; thickening adds boundary pixels in a controlled way, while pruning removes spurious spurs left over after skeletonization.

Question 9

Q.5. (a) Discuss watershed transform along with its suitable application. [8]

Accepted Answer

The watershed transform treats the (gradient) image as a topographic surface and floods it from local minima to build dams (watershed lines) that segment touching objects, mitigating over-segmentation via marker-controlled watershed; Huffman coding is a lossless compression technique that assigns shorter codes to more frequent symbols, guaranteeing exact, reversible reconstruction of the original data.

Question 10

Q.5. (a) Discuss image segmentation based on global thresholding. [8]

Accepted Answer

Global thresholding segments an image by classifying every pixel as foreground or background based on a single intensity threshold, with Otsu's method providing an automatic, optimal threshold by maximizing between-class variance; edge-based segmentation instead detects intensity discontinuities using gradient or Laplacian operators and links the resulting edge pixels into closed object boundaries.

Digital Image Processing