1 | P a g e EE5351 – DIGITAL VIDEO CODING INSTRUCTOR: Dr. K.R. Rao Summer 2009, Final Exam Tuesday, 11 August 2009 1:00PM – 2:30 PM (1:30 Hour)2 | P a g e PART 1 (Q1) to (Q5) with [4 Points] each (Q1) Which of the following regarding JPEG-LS is true? A. Has Simpler Predictive coder than CALIC B. Performance is close to CALIC C. Performance is better than CALIC D. The initial prediction approach is a variation of Median adaptive prediction E. Prediction Errors are encoded using Golomb Codes (Q2) Which of these Facsimile Encoding Groups uses Digital Compression Methods to greatly reduce transmission times? A. Groups 1 and 2 B. Groups 3 and 4 C. Only Group 3 D. Only Group 4 E. Groups 1, 2, 3 and 4 (Q3) A 2 channel Filter bank is as shown in Fig 1 below. H0 and G0 are LPF’s; H1 and G1 are HPF’s The Equations for Perfect Reconstruction are A. G0(ω) = C. H1(ω +π) ; G1(ω) = -C. H0(ω +π) B. G0(ω) = C. H0(ω +π) ; G1(ω) = -C. H1(ω +π) C. 200( ) .( 1) . ( )g n C h n; 211( ) .( 1) . ( )g n C h n   D. 201( ) .( 1) . ( )g n C h n; 210( ) .( 1) . ( )g n C h n  3 | P a g e (Q4) Which measurement(s) is (are) used to verify the performance of lossless image codecs? A. CR (compression ratio) B. Bit rate (in bits per pixel) C. MSE (mean square error) D. PSNR (peak signal-to-noise ratio) E. Histogram of pixel difference between original and reconstructed image (Q5) Which statement(s) satisfy(s) the conditions for PR (perfect reconstruction) of a two-channel filter bank? Write answer(s) ________________ A. Reconstructed output is the scaled version of the original signal. B. Reconstructed output is the delayed version of the original signal. C. Reconstructed output has no aliasing (no phase distortion) D. The analysis low pass filter is orthogonal to the twice-shifted version of itself. E. None of the above.4 | P a g e [10 Points][Problem 1] Figure1. Labeling the neighbors of pixel X5 | P a g e [10 Points][Problem 2] The contexts in JPEG-LS are computed differently from CALIC. First, measures of differences D1, D2, and D3 are computed as follows: where T1, T2, and T3 are positive coefficients that can be defined by the user. Given nine possible values for each component of the context vector, this results in 9 x 9 x 9 = 729 possible contexts. In order to simplify the coding process, the number of contexts is reduced. Clearly Explain how the number of contexts are reduced and to what number.6 | P a g e [20 Points][Problem 3] A 512x512 input image (8 bit PCM) is decomposed into 7 unequal subbands as shown in fig.2. Assume that 8 bits are allocated to each sample in the first subband, 4 bits are allocated to each sample in subbands 2,3,4, and 2 bits are allocated to the rest. This decomposed image is sent through a 9600 bits-per-second channel in a progressive manner. Specifically, stage 1 of transmission contains information of subband 1 only, stage 2 contains information of subbands 2,3,4, and stage 3 contains information of subbands 5,6,7. How long would it take to transmit each stage of transmission and what fraction of image is sent during the same time by normal transmission (without subband coding). Figure 27 | P a g e [20 Points][Problem 4]8 | P a g e [20 Points][Problem 5] 1) Show that the symmetric low pass filter nNnhh1 is linear phase for length 8. (N=0, 1, 2, …, 7) 2) Show that high pass filter nNnnhh1)1( is also linear phase for length 8. (N=0, 1, 2, …, 7) SOLUTIONS (Q1) A, B, D, E (Q2) B (Q3) A, D (Q4) A, B (Q5) A, B, C, D9 | P a g e10 | P a g e11 | P a g e12 | P a g e13 | P a g e14 | P a g e15 | P a g