Application Of Embedded GPU In Ultrasonic Image Processing

The advantages of medical ultrasonic diagnosis technology is the harmless to human body tissues, and is now widely used and applied in clinical medical diagnosis. However,with the rapid development of electronic technology, medical ultrasonic diagnostic equipment portability has become an inevitable trend, while new technology to improve the quality of ultrasound images will continue to surge,the two kinds of demand above makes portable devices need to face the problem of shortage of hardware resources. This thesis is focused on the application of embedded GPU to ultrasonic image processing, so as to solve the problem of large computation in the back-end signal processing, so that it can give full play to the system performance.This article briefly describes the basic principle of ultrasonic imaging and introduces B type ultrasound back-end processing in all aspects. Then this article the processing part of the large amount of calculation, and determine the part which needs to use GPU parallel computing. Hardware and software platform for use in this subject and Open GL ES basic principles also has a simple introduction.For compounding imaging and digital scan conversion links of back-end signal processing needing GPU computing, this article analyzes its algorithm principle in detail, while the corresponding formula is derived and used Matlab to offline simulation testing, including analyzing the output results. Finally, detailed GPU implementation process is introduced and the results of tests are carried out in real-time data corresponding analysis. The GPU implementation process utilizes Jni corresponding interfaces to call the local Open GL program via Java.Finally, from the angle of the ultrasonic diagnosis, the image segmentation algorithm is introduced to the back-end processing. For comparison of several common segmentation, the corresponding algorithm introduces the principle of comparing the segmentation quality and using time of the Matlab simulation. Then,this article gives a detailed GPU implementation and test results of FCM segmentation algorithm.