運用高頻資訊補償之超解析度演算法研究

High Frequency Compensated Super-Resolution Algorithm

張書梵(Shu-Fan Chang)

指導教授 ： 張寶基

國立中央大學/資訊電機學院/通訊工程學系/碩士(2008年)

摘要

時至現今，已有許多重建高解析度影像的方法。這些方法主要是去除影像經內插放大時的模糊效應和適當的增加空間上的資訊，以還原其高解析度的面貌。在本研究的方法中，藉由迭代反投影法(Iterative Back Projection, IBP)的原理，並加入高頻補償模型，依據應用的不同，發展出直接高頻補償(Direct High Frequency Compensated, DHFC)和估測高頻補償(Estimated High Frequency Compensated, EHFC)兩種方法。 本論文所提的方法，除了大幅改善零次，雙線性，和三次立方等，傳統內插法所放大的影像品質外，主要在於改善IBP演算法的缺點。在一般影像測試的情形下，DHFC方法可提升約1~5.7倍的速度，並提升最終PSNR值約0.1~0.3dB；EHFC方法可提升約1~6.5倍的速度，並提升最終PSNR值約0.4~0.7dB。在文字影像測試的情形下，EHFC方法可提升約1~6.5倍的速度，並提升最終PSNR值約1.2~8.3dB。

關鍵字

影像重建 ； 超解析度 ； 反投影迭代 ； 影像內插 ； 高解析度

並列摘要

Currently, there exist many high resolution image reconstruction methods. The approach of these methods is mainly to remove the blur effect of image by interpolation enlargement and appropriately to increase the space information to restore its high resolution images. In this study, two methods, the Direct High Frequency Compensated (DHFC) and the Estimated High Frequency Compensated (EHFC), are developed based on the Iterative Back Projection (IBP) principle as well as the High Frequency Compensated Model. These methods substantially improve the image quality of reconstructed images enlarged by zero, bilinear, and bi-cubic interpolations. In addition, they accelerate the process significantly compared to IBP algorithm. In natural imaging test situation, DHFC method can increase the speed by 1~5.7 times and improve the ultimate PSNR value by 0.1~0.3dB while EHFC method can increase the speed by 1~6.5 times and improve the ultimate PSNR value by 0.4~0.7dB. Moreover, in text imaging test cases, EHFC method can increase the speed by 1~6.5 times and improve the ultimate PSNR value by 1.2~8.3dB.

並列關鍵字

image interpolation ； image reconstruction ； super-resolution ； high-resolution ； Iterative Back Projection

參考文獻

[1]Conventional Television Systems, ITU-R standard BT470-6, 1970-1998.

[4] Park S C,Park M K,Kang M G. Super-resolution image reconstruction: a technical review[J]. IEEE signal processing magazine, 2003, (5):21-36

[5] R. G. Keys, “Cubic Convolution Interpolation for Digital Image Processing”, IEEE Trans. Acoustics, Speech, and Signal Processing, vol. ASSP-29, no. 6, pp. 1153-1160, Dec. 1981.

[7] W. K. Pratt, Digital Image Processing, 2nd ed. New York: Wiley, 1991.

[8] M. Unser, A. Aldroubi, M. Eden, “B-Spline Signal Processing: Part II-Efficient Design and Applications”, IEEE Trans. Signal Processing, vol. 41, pp. 834-848, Feb. 1993.