运动模糊图像实时恢复的多DSP方案

为解决相机与被拍摄物体之间存在较大相对运动而产生的图像模糊,提出使用三个定点DSP和一个浮点DSP共同完成运动模糊图像实时恢复的方法。研究运动模糊图像的退化模型,采用维纳滤波方法进行运动模糊图像恢复。阐述了定点DSP完成二维FFT的原理和注意事项,说明了四个DSP在图像恢复过程中各自的功能。对于512×512大小的图像,定点DSP完成二维FFT的时间为17.7ms,浮点DSP完成浮点数据乘法和除法的时间总共为32.51ms,据此计算了各个DSP的工作负荷。通过这些指标说明,此硬件结构可以完成512×512大小运动模糊图像的实时恢复。

Real-time Restoration Research of Motion Blurred Image Based on DSPs

T o absolve imag e blur phenomena caused by r elat ive mo tion betw een camera and o bject s, a realt ime blur image restor at ion metho d is put forw ard, w hich is composed of three fix ed po int DSP and o ne float point DSP. Degenerat ion model is resear ched. From the point of v iew of sof tw are, Wiener filtering is adopted to restore image. T w o-dimensional FFT principle and impo rtant quest ions ar e presented executed by fix ed DSP, the respect ive functio ns of four DSP is ex plained during the co urse o f imag e restor at ion. For the image of 512×512, the t ime o f two-dimensional FFT by fix ed DSP is 17. 7 ms, the gro ss t ime of f loat po int mult iplicat ion and division by flo at point DSP is 32. 51 ms, according to above t ime, the load of ever y DSP is calculated. Fr om these data, the hardw are can complete real-t ime r estorat io n of 512×512 image.