本論文提出一種創新的彩色影像壓縮演算法,用來實現低複雜度的影像壓縮積體電路設計。為了達到低複雜度的需求,捨棄方塊編碼(BTC)改用絕對值動量保留區塊截短碼(AMBTC)進行區塊壓縮,以平均值取代標準差的運算,並用已知的位元圖取代區塊影像值。熵編碼部分使用哥倫布萊斯編碼,將較常使用之位元圖分配較短的碼做儲存。由於方塊編碼類的壓縮方法會導致影像的馬賽克化,因此本論文根據影像高低頻做壓縮區塊大小的調整,搭配像素預測降低高頻處的失真率。本論文提出的「基於採樣模塊分類及絕對值動量保留區塊截短碼之彩色影像壓縮演算法」與以往研究相比,在相近的影像品質下,比較於先前文獻,本論文所提出技術之壓縮率可提高25%以上,且PSNR能維持在30 dB左右。
A novel color image compression algorithm is proposed for very large-scale integration (VLSI) circuit design in this thesis. To achieve the target of low complexity, this thesis uses Absolute Moment Block Truncation Coding (AMBTC) instead of Block Truncation Coding (BTC). This thesis also replaces standard deviation with average and substitutes the known bitmap for all values in a block. The bitmap which has the appearance of high-frequency will be stored as fewer bits by using Golomb-Rice coding. This thesis analyzes the frequency of image and then uses prediction to decrease mosaic. Compared with previous studies, the experimental results show that almost 25% compression ratio can be increased by using the algorithm in this thesis with similar PSNR. The average of PSNR values is about 30 dB.