Image Compression Method And Real-time Implementation For Capsule Endoscope System

As a painless,non-invasive,safe and comfortable way to diagnose gastrointestinal diseases,capsule endoscopy systems have been widely used in domestic and international clinics.After the subject swallows the capsule,the built-in high-definition camera transmits the image data out of the body through a wireless transmission device,thus making a diagnosis of the condition.With the increasing requirements of image resolution and frame rate for accurate clinical diagnosis,the contradiction between system transmission data volume and limited channel bandwidth becomes more and more prominent,and image compression technology becomes the key to improve the data transmission performance and reduce the power consumption of the system.The research of this thesis involves the design and implementation of capsule endoscopy image compression and transmission system,which faces two major challenges: high-fidelity compression and strong real-time compression.Based on the characteristics of capsule endoscope images,this thesis studies compression methods,designs hardware encoder architecture,and develops real-time compression and transmission system.For compression method,this thesis analyzes capsule endoscopy CFA(Color Filter Array)image characteristics,selects the lossless compression algorithm HPCM(Hierarchical Prediction and Context Modeling)as the base algorithm,and proposes a lightweight prediction and coding scheme for the problem of high computational complexity of hierarchical prediction and arithmetic coding.A low-complexity intra-frame lossless compression algorithm MHPCM is integrated,and the experimental results show that the compression ratio loss is less than 5%.Further,this thesis analyzes the problem of low efficiency of continuous tone image lossless compression standard JPEG-LS for CFA image prediction and proposes a prediction template reconstruction scheme;combined with the correlation between image sequences,this thesis proposes a motion compensation-based inter-frame prediction scheme.An inter-frame lossless compression algorithm MJPEG-LS is integrated for CFA image,and the experimental results show that the compression ratio is improved by more than 60%.For encoder architecture,to address the computational characteristics and the non-causal template implementation bottlenecks of the low-complexity intra-frame lossless compression algorithm MHPCM,this thesis proposes a micro-block rearrangement scheme,and designs a fully-pipelined hardware encoder architecture.The compression ratio of the inter-frame lossless compression algorithm MJPEG-LS is high while the computational complexity of motion compensation is also very high.Therefore,this thesis proposes some optimization schemes,such as the simplified block data addressing method based on the exchange binary bits,the high-data-reuse full search block matching method based on the sliding window structure,and designs a globally-pipelined and partially-parallel encoder architecture to improve the encoding rate and reduce the hardware overhead.For system implementation,this thesis designs a miniaturized and low-power image compression and transmission system based on FPGA.After verification,the MHPCM encoder achieves a size of 93 K equivalent gates,a lossless compression ratio of 1.74,and a power consumption of only 6.17 mW,which is suitable for capsule endoscopy applications with low frame rate image requirements;the MJPEG-LS encoder achieves a size of 267 K equivalent gates,a lossless compression ratio of 2.44,and a power consumption of only 7.56 mW,which is suitable for capsule endoscopy applications with higher frame rate image requirements.