Lossless Medical Image Compression Based On Region Segmentation And Regression Model

The widespread use of digital medical systems has led to a dramatic increase in the number of digital medical images.Medical images are an important diagnostic tool for doctors and therefore have higher resolution,which results in more space for the storage of medical images.On the one hand,the amount of image data is exploding,and on the other hand,limited bandwidth and storage space are difficult to expand in a short period of time,which brings great pressure for image storage and transmission.Therefore,an effective compression method for medical image design is not only capable of reducing storage pressure but also reducing transmission time,which is of great significance.In this paper,a medical image lossless compression method based on image segmentation and regression model is proposed.Based on the idea of Differential Pulse Code Modulation(DPCM),image segmentation and regression algorithms are combined for data de-correlation.The de-correlated data is entropy encoded to achieve compression of the data set.Image segmentation is the first step of the compression method proposed in this paper-Based on the density and anatomical properties of medical images,a new image segmentation method is proposed.The method adopts a simple and complicated method,and according to the density characteristic of the image,first divides the region of the medical image that has a large difference with the surrounding tissue density threshold,and for the region with similar density threshold in the image,through the anatomical model,in the image.The segmentation candidate region is obtained,and then each candidate region is segmented based on the density feature.Finally,the whole image is divided into 12 regions.The experimental results prove that the segmentation method proposed in this paper is effective and can provide conditions for constructing the prediction model in the next step.One of the key challenges in obtaining effective compression is to generate accurate prediction models.In the construction of predictive models,multiple linear regression,polynomial regression,decision tree regression and BP neural network algorithms are analyzed,and appropriate optimization methods are selected according to experimental results.And the network structure,and strive to maximize the effectiveness of each algorithm in the training model process.Through experimental comparison,decision tree regression and multiple linear regression are selected as the predictor algorithm for training corresponding regions with the advantages of high prediction accuracy and fast training speed,and finally the adaptive prediction model is obtained.During the compression process,the corresponding region prediction model is adaptively selected for the current pixel according to the reference information,and the decorrelation operation is completed.After the prediction model is constructed,the de-correlated data is encoded using the best entropy encoder to complete the entire compression process.Finally,using the compression method proposed in this paper,20 sets of test data are compressed,and the final compression result is compared with the compression result of the traditional compression method.The experimental results show that the compression effect proposed in this paper is better than JPEG2000,which is better than the least square method and PPMD.The method is improved by 28.18%,which proves that the compression method proposed in this paper is effective for the compression of a large number of medical images.