一個植基於顏色替代的高效能與高效率影像壓縮技術

現今資訊科技與網際網路的發達，人們常常傳輸大量多媒體資料，隨之而來的問題是，影像不僅佔用大量的儲存空間，在傳輸的過程中也花費了許多的時間。因此，如何以最快的速度傳輸與利用最小的空間儲存影像資料是個極大的挑戰，而「影像壓縮技術」也就顯得更加重要。一張全彩的影像可能涵蓋許多重複的顏色或是不重要的顏色。而在影像壓縮過程中，LBG會因為重複訓練相同的顏色而導致花費了更多的時間成本；H and F雖然執行時間快速，但其壓縮品質會因為Nb值的設定而有所影響。因此，為了解決此一問題，本論文提出了一個名為CRLBG演算法，其主要目的是將LBG與H and F兩者做截長補短的改良，將LBG的高壓縮品質與H and F的低時間成本概念合而為一。本研究主要概念為在產生初始編碼簿以前先進行動態選擇顏色替代的元素，此動態選擇是根據該像素的R、G、B值做判斷，判斷哪個元素進行顏色替代後會使MSE值越小，則選擇該元素做顏色替代訓練，將原本影像中相似的顏色藉此一動作調整為相同顏色，減少了壓縮過程中像素樣本的數量，使得CRLBG演算法對於初始編碼簿的設計及收斂可以更為快速，且取得代表色的機率也上升許多，所以CRLBG演算法在時間成本與PSNR值上相較於LBG、SOM與LazySOM皆有不錯的表現。

關鍵字

影像壓縮；色彩量化；顏色替代； LBG

並列摘要

Under the development of advancement in information technology and Internet, nowadays, people often use computers, mobile phones, tablets and other devices to transmit large number of multimedia data, thus it causes image not only takes up a lot of storage space, but also spend a lot of time in the process of transmission. Therefore, how to use the fastest speed of transmission and minimal storage space to handle image data is a big challenge. Consequently, image compression technique will become more popular and important. A full-color image may contain many repetitive or unimportant colors. During the process of image compression, LBG may repeatedly train the same colors to lead to spend more time cost; although H and F algorithm take not much time, the value of Nb may affect the quality of compression. In order to solve these image compression problems, the thesis proposes a new approach called CRLBG algorithm, which main purpose is to combine LBG algorithm's high quality of compression and the low time cost of H and F's algorithm to attain improvement. The main concept of this thesis is as follows: before generating the initial codebook, dynamically select element of RGB (color model) to perform "Color Replacement", this dynamically select based on the pixel R, G, B value judgment to determine which color alternative will make an element MSE smaller values to perform "Color Replacement", which makes the original similar color to alter same colors and then it may reduce the number of pixel samples. This step makes an initial codebook design, and convergence can be conducted more quickly. Accordingly, the color may increase significantly. According to the simulation results, the proposed CRLBG algorithm outperforms LBG, SOM and LazySOM in image compression.

並列關鍵字

Image Compression ； color Quantization ； Color Replacement ； LBG

參考文獻

[3] 吳佳珉，ELBG：植基於LBG的新彩色影像壓縮技術，國立屏東科技大學資訊管理學系碩士論文，2012。

[7] Nian, Y., He, M., Wan, J., “Lossless and near-lossless compression of hyperspectral images based on distributed source coding,” Journal of Visual Communication and Image Representation, Vol. 28, pp. 113-119, 2015.

[11] Abouali, A.H., “Object-based VQ for image compression,” Ain Shams Engineering Journal, Vol. 6, No. 1, pp. 211-216, 2015.

[12] Hu,Y. C. and Lee, M. G., “A K-means Based Color Palette Design Scheme with the Use of Stable Flags,” Journal of Electronic Imaging, Vol. 16, No. 3, 033003, 2007.

[13] Kohonen, T., “Self-organizing map,” Berlim, 1995.

國際替代計量

一個植基於顏色替代的高效能與高效率影像壓縮技術

全文下載

主題瀏覽