Design And Optimization Of Printer Image Processing Engine Based On Loongson 2K1000B

The printer image processing engine is used to process the image decoding of files to be printed.Printer image decoding includes JBIG decoding and JPEG decoding.More than half of the processor's execution time is occupied by image decoding when printing files.Image decoding speed is one of the main bottlenecks of the printer,which directly affects the printing speed of the printer,and it is critical to increase the image decoding speed.JBIG/JPEG decoding function library and Loongson 2K1000 B integrated VPU(Video Processing Unit)are used to implement the first version of printer image processing engine,which includes JBIG decoder and JPEG decoder.The JPEG decoder consists of a software decoder and a VPU-based decoder.Then the thesis obtains the performance bottleneck,hotspot functions and optimization methods of the image processing engine through evaluation and analysis.Then the image processing engine is optimized.For the JBIG decoder,the process of the main function is optimized;then the memory copy is optimized by using the Loongson extended memory access instruction;Finally,the resolution control strategy is adopted.When decoding an image exceeding 600 PPI(Pixels Per Inch),the decoding is ended in advance at the 600 PPI level.For the JPEG software decoder,firstly,Huffman decoding is optimized by constructing a short word code table;then multimedia instructions are used to optimize inverse quantization and inverse discrete cosine transform;Finally,the color space conversion is optimized by skipping the intermediate upsampling process and the second table lookup.For the VPU-based decoder,the optimized memory copy is also used to improve the performance.The optimized image processing engine is tested on the Longson 2K1000 B platform.The result shows that the optimized image processing engine in this thesis can meet the actual needs of a domestic printer.For the 600 PPI A4 black-and-white image,the optimized JBIG decoder consumes 16.6% less time than before,and it takes 1892 ms.For the 600 PPI A4 color image,the optimized JPEG software decoder consumes 29.5% less time than before,and it takes 1822 ms.The result of color image decoding shows that the optimized JPEG software decoder is only slower than the VPU decoder when decoding high-pixel images whose pixels are larger than 10 million.Finally,it is proposed to combine the optimized software decoder and VPU-based decoder to achieve the final JPEG decoder,which ensures shorter decoding time under different pixels.