Study On JPEG XR Encoding Algorithm And Implementation Of LBT On FPGA

Images are the important carrier for people to transport information and communicate each other.With the progress of social and science-technology,the requirement of people for quality of images is becoming higher and higher.Especially,in some areas such as aeronautics-astronautics,video security,telemedicine and high-definition-television,it is more important for the image quality on transmission and storage.Therefore,at the same time of improving bandwidth technology,it has not been stopped to study on image compression.Most of all,the JPEG family has developed largely which from JPEG to JPEG 2000 that has a lot of efforts of the researchers.In 2009,JPEG XR was been released officially.The complexity of lapped bi-orthogonal transform(LBT)of JPEG XR is same as the discrete cosine transform(DCT)of JPEG.However,the quality of decoding image of JPEG XR is comparable with the JPEG 2000 which uses the more complicated discrete wavelet transform(DWT).For those reasons above,JPEG XR has been studied widely and a good application prospect.LBT of JPEG XR has 4 subordinate transforms.The 2 photo overlap transforms are to eliminate block effects of images,and the 2 photo core transforms are to transform images from spatial domain to frequency domain.LBT provides data which in the format based on frequency distribution.It improves the encoding efficiency and decoding image quality of JPEG XR greatly,so LBT is the very important part of JPEG XR.Because LBT has lots of multiplications and divisions,it needs high-quality hardware to implement.FPGA has many advances,for example,the high-frequency,the strong parallel processing capability and so on.Therefore,this design uses the powerful signal drive-ability of FPGA fully,and it replaces of the linear-structure multiplier with the mixed state machine.The design reduces the FPGA hardware source consumption that LBT needs.It also uses single RAM cycle structure to store the intermediate transform result of LBT.Through cross switch of RAM channel controller to perform the operations of reading and writing between the 4 subordinate transforms of LBT and RAM.And that saves large storage space of FPGA.This paper from 3 aspects that search and analysis on JPEG XR encoding algorithm,simulation on MATLAB,design and implementation on FPGA to discuss.Firstly,it mainly analyzed LBT algorithm,quantization,adaptive prediction,adaptive scan and adaptive entropy coding of JPEG XR.The most important characteristic of JPEG XR is that it can choose dealing way in real time according to image contents,it is the adaptive coding.Secondly,for the RAM storage structure of FPGA that is one-dimensional address-data format,it used MATLAB to simulate and verify those processes before entropy coding.Then,it used FPGA to design LBT,in the design,the mixed state machine was used to realize the cross-communication between each transform algorithm and RAM controller of LBT subordinate transforms.Last,it used CP2012 to put out the result of LBT to the serial debugging assistant in upper computer,and compared it with the LBT transform result on MATLAB.The test results show that the LBT result on FPGA is almost same to it on MATLAB.The transform function has been implemented by this design.