隨著醫療科技的進步,「醫學影像」大量地被使用在臨床醫學上。為了讓醫療儀器與電腦間能夠相互溝通使資訊可以交換查閱,並進行影像分析處裡,進而發展出了一套醫療數位影像傳輸協定(Digital imaging and communications in medicine;DICOM)。醫學影像著重完整呈現最原始的資料,每筆資料量甚為龐大,因而衍生了儲存空間的問題。而透過資料壓縮的方式便可達到節省空間的目的。然而為了保有資料的完整性,必須採取所謂非失真的壓縮方法。為此,本研究提出了一個高效能的演算法來進行醫學視訊的非失真壓縮。在空間域的壓縮上,本研究以預測編碼為基礎,並在遭遇影像邊界時以最小平方法進行預測器係數的修正,使預測器能具備良好的預測效能與運算複雜度。而在時間域方面的壓縮上則採用快速區塊移動估測法。透過相鄰影像間的關聯性便可決定當前影像該採用空間域的編碼法或是時間域的移動估測法。經由實驗證明本研究所提出之演算法不論在空間域或時間域兩方面都能有良好的表現,因此能夠提升醫學影像壓縮的效能。
With the advances in medical technologies, medical imaging has been widely applied in clinical medicine. In order that the captured images can be exchanged and analyzed between the medical equipment and the host computer, a so-called digital imaging and communications in medicine (DICOM) protocol is developed for the transfer of digital medical images. On the other hand, a large storage capacity is usually required for medical images for its large dimension and high resolution. Therefore, lossless image compression technique is usually required for efficient storage of medical images and sequences. Aimed to provide an efficient algorithm for the compression of medical image sequences, we propose in this paper a coding scheme that can switch between two prediction modes; the intra mode coding scheme and the inter mode coding scheme. For intra mode coding, a Least-squares-based adaptive predictor is applied for the removal of spatial redundancy. Moreover, the predictor coefficients are adapted whenever an edge or a boundary is detected so that a large prediction error and the high computational cost of LS adaptation process can be avoided. For inter mode coding, a fast motion estimation algorithm is applied for the removal of temporal redundancy. Moreover, the decision of using an intra or inter mode is mainly based on the correlation between consecutive image sequences. Experimental results show that the proposed algorithm can have a very good compression ratio as well as a very good run-time performance, which justifies the usefulness of the proposed approach for the compression of medical image sequences.