(SEM VI) THEORY EXAMINATION 2022-23 IMAGE PROCESSING

IMAGE PROCESSING (KCS-062)

Important Questions & Answers – Section-wise

SECTION A

(Attempt all questions – 2 × 10 = 20 marks)

(a) Advantages of Wiener filter

The Wiener filter provides optimal restoration by minimizing the mean square error between the original and restored image. It considers both noise and image degradation, making it more robust than inverse filtering in noisy environments.

(b) Digital Image Processing

Digital Image Processing is the use of computer algorithms to manipulate digital images for enhancement, restoration, analysis, and interpretation.

(c) PSF (Point Spread Function)

PSF describes how a point source of light is spread by an imaging system. It characterizes image blurring caused by the system.

(d) Erosion and Dilation

Erosion removes pixels from object boundaries, shrinking objects, while dilation adds pixels to boundaries, expanding objects. Both are morphological operations.

(e) Degradation

Degradation refers to the distortion or deterioration of an image due to factors like noise, motion blur, or defocus during image acquisition.

(f) Reflectance

Reflectance is the proportion of incident light reflected by an object’s surface, which determines image intensity values.

(g) Inverse filter vs Wiener filter

Inverse filtering ignores noise and amplifies it, while the Wiener filter accounts for noise statistics and degradation, producing better results in real conditions.

(h) Use of Boundary Extraction

Boundary extraction identifies object outlines, useful for shape analysis, object recognition, and segmentation.

(i) Image enhancement vs restoration

Enhancement improves visual appearance subjectively, whereas restoration aims to recover the original image using mathematical models.

(j) Harmonic mean filter

The harmonic mean filter reduces salt noise and preserves edges better than arithmetic mean filters.

SECTION B

(Attempt any THREE – 10 × 3 = 30 marks)

Q2(a) Image processing steps with block diagram

Image processing includes image acquisition, preprocessing, enhancement, segmentation, feature extraction, classification, and interpretation. Each stage transforms the image to extract meaningful information.

Q2(b) Piecewise linear transformations

Piecewise linear transformations enhance contrast by mapping input intensity ranges to output ranges. Examples include contrast stretching and gray level slicing.

Q2(c) Band pass filter technique

Band pass filters allow a specific range of frequencies to pass while attenuating others. They reduce noise by removing very high and very low frequencies.

Q2(d) Watershed segmentation algorithm

Watershed segmentation treats the image as a topographic surface. Regions are formed by flooding from local minima until boundaries meet, separating objects effectively.

Q2(e) Data compression and run length encoding (RLE)

Compression reduces storage and transmission costs. RLE compresses data by encoding consecutive identical pixels as a single value and count.

SECTION C

Q3(a) Low, mid, high-level processing; sampling & quantization

Low-level processing includes noise removal and enhancement. Mid-level involves segmentation and feature extraction. High-level processing focuses on interpretation. Sampling converts continuous images to discrete grids, while quantization assigns intensity levels.

Q3(b) Correlation vs convolution

Correlation measures similarity without reversing the filter, while convolution reverses the filter before operation. Convolution is fundamental in filtering.

Q4(a) Frequency domain filtering

Images are transformed using Fourier Transform. Low-pass filters remove high frequencies (smooth images), while high-pass filters enhance edges.

Q4(b) Bit plane slicing and homomorphic filter

Bit plane slicing analyzes individual bit contributions to image quality. Homomorphic filtering separates illumination and reflectance to enhance contrast.

Q5(a) Image restoration and block diagram

Image restoration recovers the original image from a degraded one using a degradation model and noise estimation.

Q5(b) Median and midpoint filters

Median filter replaces a pixel with the median of neighbors, removing impulse noise. Midpoint filter averages maximum and minimum values, reducing uniform noise.

Q6(a) Edge detection and edge linking

Edge detection finds intensity discontinuities. Edge linking connects detected edges into continuous boundaries. Detection identifies edges; linking organizes them.

Q6(b) Image segmentation and conditions

Segmentation partitions an image into regions such that pixels within a region are homogeneous, and regions are distinct.

Q7(a) JPEG vs MPEG

JPEG compresses still images using DCT, while MPEG compresses video using temporal redundancy and motion estimation.

Q7(b) Image compression vs recognition

Compression reduces data size; recognition identifies and classifies objects within images.