在JPEG2000系統中，EBCOT需要相當大的運算量，為了增進EBCOT的效率，我們提出一個快速演算法Pass Propagation。此演算法預測下一個位元平面的pass狀態，再利用相鄰位元以傳播的方式得到下一個位元平面所有pass狀態。透過狀態的傳播，並且考慮位於邊界的位元作演算法的調整，減少掃描次數同時可以得到正確的掃描結果。根據實驗結果，我們提出的快速演算法可以減少灰階影像42%到52.7%的掃描時間；減少彩色影像11.8%到34%的掃描時間。相較於之前提出的演算法有更高的效能。 我們將JPEG2000編碼器實現於DSP平台上，利用DSP在數位訊號處理的優勢，透過程式與設定的最佳化並加入快速演算法，應用多項加速方法後可以使影像的編碼時間從23.25秒/張減少到2.71秒/張。最後利用包含ARM和DSP的雙核心平台將JPEG2000編碼器應用為一個拍照系統，使用者可依所需的影像品質或壓縮率來決定編碼影像大小。 In JPEG2000, embedded block coding with optimized truncation (EBCOT) requires a lot of computations. To enhance the performance of EBCOT, an efficient EBCOT algorithm is proposed. This algorithm consists of pass-prediction and pass-propagation. Pass prediction predicts the pass state in the next bit-plane. In particular, pass propagation generates the state of neighboring passes in the next bit-plane. This work also accomplishes boundary pass classification. With the propagated state, the pass state of the code-block can be classified for indicating the correct coding method. Various simulations have been conducted. The experimental results show that the proposed algorithm reduces the number of scan cycles by 42% to 52.7% for gray-level images and 11.8% to 34% for color images. The comparison demonstrates that the performance of our proposed method is higher than that of previous methods. We implement this algorithm on a digital signal processing platform, with optimization of program and configuration, the encode time of encode images reduced from 23.25 second/per image to 2.71 sec/per image on the platform. The JPEG2000 encoder is applied to a photograph system, user can decide encoded file size according to the image quality or compression ratio which user selects.
