Computer Vision

Daffodil International University Faculty of Science & Information Technology Department of Computer Science & Engineering Midterm Examination, Fall 2026 Course Code: CSE437, Course Title: Computer Vision Level: 4 Term: 3 Batch: 64 Time: 01:30 Hrs Marks: 25 Answer ALL Questions [The figures in the right margin indicate the full marks and corresponding course outcomes. All portions of each question must be answered sequentially.] 1. a) Given the following 5 × 5 grayscale image: I = [10 10 10 100 100 10 20 20 110 110 10 20 30 120 120 10 20 20 110 110 10 10 10 100 100] Apply histogram equalization to the given grayscale image, quantizing the output into 16 intensity levels. Show the gray-level mapping and construct the equalized image matrix. [5] 2. a) For the following 5 × 5 image I, calculate the Gaussian-filtered value at the center pixel using the given 3 × 3 kernel (with Reflect padding). I = [10 10 10 100 100 10 20 20 110 110 10 20 30 120 120 10 20 20 110 110 10 10 10 100 100] [2] b) Using the center 3 × 3 region of the ORIGINAL image provided in Question 2(a), calculate G_x and G_y with the Sobel masks. Then determine the gradient magnitude and gradient orientation θ = atan2(G_y, G_x), in degrees. Treat both masks as correlation masks (without kernel flipping). [3] 3. a) For an RGB pixel (120, 180, 60), Determine C_max, C_min and Δ. [1] b) Convert the pixel to HSV by calculating Hue, Saturation and Value. Then convert it to HSI by calculating Hue, Saturation and Intensity using the standard HSI model. Express both Hue values in degrees and all other values in [0, 1]. [4] 4. a) An 8-bit image has the following histogram: Intensity: 20 | 60 | 100 | 140 | 180 Frequency: 4 | 5 | 3 | 6 | 2 Calculate the normalized probabilities and the global mean intensity. [1] b) Using Otsu's method, analyze candidate thresholds T = 20, 60, 100 and 140. For each T, determine class weights (w_0, w_1) and class means (μ_0, μ_1). Calculate the between-class variance σ^2_B = w_0 w_1 (μ_0 - μ_1)^2 and justify the selection of the optimal threshold. Assume intensity values ≤ T belong to Class 0. [4] 5. a) Given query descriptor Q = [2, 4, 6, 8] and candidate descriptors D_1 = [3, 5, 7, 9], D_2 = [4, 6, 8, 10] and D_3 = [7, 9, 11, 13], compute all three Euclidean distances. [2] b) Identify the nearest and second-nearest candidate descriptors. Apply Lowe's ratio test using τ = 0.75, and determine whether the nearest match is accepted or rejected, and justify your decision. Compute the Hamming distance between B_1 = 10110110 and B_2 = 11010011. [3]