【多媒體概論】期中記錄

作者：Dr.Zero│2019-04-12 01:31:28│巴幣：0│人氣：607

答案僅供參考，因為我也不會。

Q1.The histograms of three imahes are illustrated below. For each image, sketch a transformation function in the figure below that will help to equalize the histogram. Explain briefly your reasoning.

Fig.a

Fig.b

Fig.c

A1. 答案為各圖表積分的結果。

Q2. What are two phases of data compression?

A2. 取樣、量化。

Q3. Construct a Huffman tree for the following symbol set A, where P(ai) represent the probability of ai.

* A={a1, a2, a3, a4, a5}

* P(a1) = 0.2, P(a2) = 0.4, P(a3) = 0.2, P(a4) = 0.1, P(a5) = 0.1

A3.

a1 = 001, a2 = 1, a3 = 01, a4 = 0000, a5 = 0001

Q4. What is the average length of the above constructed tree? By any chance, could this value be smaller than its entropy?

Q5. Given a Huffman coding is not working well, how could we further improve this coding? For your proposed improposed idea, is there any drawback (if so, please explain)?

A5. 不只計算a1~a5的機率，也計算複合型的機率(ex:{a1a1,a1a2}等)，可提高壓縮率。缺點是提高計算複雜度。

Q6. In terms of transforms, we mentioned Discrete Karhunen-Loève transform (KLT) and Discrete cosine transform (DCT) in the class, please describe their pros and cons as compared with each other.

Q7. Please utilize the following figure to briefly explain the idea of transform coding.

Q8. For video compression, what is the major difference as compared to image compression algorithms?

Q9. Given an image X, assume we reduce X's contrast into 1/2 and save as Y. We then restore Y's contrast by increasing by increasing its contrast two times and save as Z. Does image Z look like X?

A9. 相似但不完全相同，因為有數值運算的誤差。舉個例子：X的其中兩個原始資料是20和21，變換Y會變成10和10，還原到Z就變成20和20了。

Q10. Among RGB three color channels, which channel are human beings feel more sensible? Please briefly explain your answer.

A10. 綠色最敏感，原因是人類的視錐細胞中，接收RGB的數量比例大概為2:4:1。

Q11. What is "Afterimage Effect"?

A11. 人類一直看同樣顏色時，會對該顏色進行補正強化，也因此一直看負片處理的照片後，突然切換成黑白的照片，會看到一瞬間的原圖。

Q12. Please explain "The Uncertainty Relation".

Q13. Please briefly explain brighness perception via the following figure.