Research And Implementation Of Lossless Compression Of 16-bit CT Images Based On JPEG

The problems of medical image transmission and preservation become increasingly important in the Picture Archiving and Communication System which is used widely. To achieve the requirement of the storage and rapid transmission, medical images must be compressed because of the large amount of data. When compressing medical images, detail informations must be kept to ensure the accuracy of medical diagnosis. So the researches of medical image coding algorithm are currently concentrated in lossless compression, nearly lossless compression, and lossless compression in region of interest. Among all the coding algorithms, the core of the encoder is Huffman coding.When compressing image with Huffman coding, the image data needs to be scanned twice to establish the Huffman code table, which greatly reduces the encoding speed.8 bits image Huffman coding tables are given in the JPEG standard, so the process of scanning data can be omitted in the process of encoding.But the JPEG standard does not give the typical Huffman code tables of 12 bits and 16 bits image. Currently the process of 16 bits medical image with Huffman coding still needs to scan the image data.In order to solve the above problems, this paper presents a common Huffman code short table for the 16 bits DICOM medical image. which can improve the encoding speed. Firstly, study the DICOM image format and read DICOM image pixel information. And then count a large number of 16 bits DICOM images to obtain the probability model of 16 bits image symbol. In order to improve the reliability of the statistics, the Good-Turing Estimate is adopted to recalculate the symbol probability to solve the zero probability event and smooth probability. Finally study Huffman coding short table generating method and create common Huffman coding short table with above statistics. When encoding and decoding image data, the Huffman coding table can be directly generated by short code table, and the process of scanning the image data is removed, which improves the encoding speed.Design control experiments to compare the performance of the common Huffman code short table in this paper and that of the optimal coding table which generated in general method. Experimental results show that the common and short Huffman code table created in this paper for the 16 bits DICOM image can improve the encoding speed increased by 40%, and keep the same compression ratio almostly compared with the optimal coding table.