-
摘要:
传统全聚焦图像融合以相机多次曝光拍摄的多聚焦图像为基础,光场相机在单次曝光后可计算空间任意深度的重聚焦图像,为后期全聚焦图像的获取提供便利。提出了一种基于小波变换的光场全聚焦图像获取算法,可有效避免传统空域图像融合算法的块效应,获得较高质量的全聚焦图像。该算法通过对微透镜阵列光场相机获得的4D光场数据进行空间变换与投影,得到用于全聚焦图像融合的重聚焦图像,对各帧重聚焦图像进行小波分解提取高、低频子图像集,提出区域均衡拉普拉斯算子、像素可见度函数分别构建融合图像的高、低频小波系数实现图像融合,其性能优于传统的区域清晰度评价函数。实验验证了所提算法的正确性和有效性,采用Lytro光场相机的原始数据计算了融合全聚焦图像,与传统图像融合算法相比,人眼视觉效果更好,客观图像指标也得到了提高。
Abstract:Traditional all-in-focus image fusion based on the multi-focus images which are captured by multiple-exposure of the camera. Light field camera has the ability of calculating the refocused images at any depth after a single exposure, which makes it more advantageous in all-in-focus image calculation. A light field all-in-focus image fusion method based on wavelet transform is proposed in this paper. Compared with the spatial image fusion method, the proposed method can effectively avoid the block artifacts and obtain a fused image with high quality. First, the refocused images used for the all-in-focus image calculation can be computed through shearing and projecting the 4D light field captured by the microlens-based light field camera. Then, the wavelet transform are applied to the refocused images and the high-frequency and low-frequency sub-images are extracted respectively. Finally, the balanced Laplace operator and pixel visibility function are proposed to evaluate the sharpness of the sub-image and to get a high-quality fusion image.Compared to the traditional region based sharpness evaluation function, the proposed method has a better performance. The experiment results prove the correctness and validity of the proposed method. The raw images captured by Lytro light field camera are used to calculate the all-in-focus image. Compared with the traditional image fusion methods, the visual effect is better and the quantitative indices are also improved with the proposed method.
-
Key words:
- light field camera /
- digital refocusing /
- all-in-focus /
- wavelet transform /
- Laplace operator /
- local sharpness /
- image visibility
-
表 1 Flower样本图像不同融合算法性能指标比较
Table 1. Comparison of performance indices of different fusion algorithms based on image Flower
算法 E AG FD EI Sobel算法 6.867 6 6.247 0 6.899 1 66.534 0 Prewitt算法 6.863 4 5.832 6 6.327 0 62.642 0 Laplace算法 6.883 0 6.866 8 7.683 7 72.307 3 本文算法 6.889 6 7.005 5 7.849 8 73.720 3 表 2 Forest样本图像不同融合算法性能指标比较
Table 2. Comparison of performance indices of different fusion algorithms based on image Forest
算法 E AG FD EI Sobel算法 5.754 4 2.532 8 2.913 6 26.515 7 Prewitt算法 5.749 2 2.290 5 2.576 6 24.273 5 Laplace算法 5.801 1 3.001 8 3.523 5 31.013 4 本文算法 5.809 9 3.087 5 3.630 5 31.903 3 表 3 Zither样本图像不同融合算法性能指标比较
Table 3. Comparison of performance indices of different fusion algorithms based on image Zither
算法 E AG FD EI Sobel算法 6.293 5 4.467 5 5.186 5 48.385 4 Prewitt算法 6.269 5 4.018 4 4.518 2 43.756 6 Laplace算法 6.271 6 4.864 9 5.677 3 52.447 4 本文算法 6.298 7 4.942 5 5.794 0 53.150 1 -
[1] NG R.Light field photography with a hand-held plenoptic camera: CSTR 2005-02[R].Stanford: Stanford University, 2005. [2] NG R.Fourier slice photography[J].ACM Transactions on Graphics, 2005, 24(3):735-744. doi: 10.1145/1073204 [3] TAO M W, HADAP S, MALIK J, et al.Depth from combining defocus and correspondence using light-field cameras[C]//IEEE International Conference on Computer Vision.Piscataway, NJ: IEEE Press, 2013: 673-680. [4] BOK Y, JEON H G, KWEON I S.Geometric calibration of microlens-based light-field cameras using line features[C]//European Conference on Computer Vision.Berlin: Springer, 2014: 47-61. [5] YOON Y, JEON H G, YOO D, et al.Light-field image super-resolution using convolutional neural network[J].IEEE Signal Processing Letters, 2017, 24(6):848-852. doi: 10.1109/LSP.2017.2669333 [6] LIU J G.Smoothing filter-based intensity modulation:A spectral preserve image fusion technique for improving spatial details[J].International Journal of Remote Sensing, 2000, 21(18):3461-3472. doi: 10.1080/014311600750037499 [7] Li S T, KANG X D, HU J W.Image fusion with guided filtering[J].IEEE Transactions on Image Processing, 2013, 22(7):2864-2875. doi: 10.1109/TIP.2013.2244222 [8] FEICHTENHOFER C, FASSOLD H, SCHALLAUER P.A perceptual image sharpness metric based on local edge gradient analysis[J].IEEE Signal Processing Letters, 2013, 20(4):379-382. doi: 10.1109/LSP.2013.2248711 [9] QIN X, ZHENG J, HU G, et al.Multi-focus image fusion based on window empirical mode decomposition[J].Infrared Physics & Technology, 2017, 85:251-260. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=f7a841d59be2b54531b9fa2b7fe14293 [10] FAKHARI F, MOSAVI M R, LAJVARDI M M.Image fusion based on multi-scale transform and sparse representation:Image energy approach[J].IET Image Processing, 2017, 11(11):1041-1049. doi: 10.1049/iet-ipr.2017.0104 [11] NAYAR S K, NAKAGAWA Y.Shape from focus[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994, 16(8):824-831. doi: 10.1109/34.308479 [12] DENG C, MA W, YIN Y.An edge detection approach of image fusion based on improved Sobel operator[C]//IEEE International Congress on Image and Signal Processing.Piscataway, NJ: IEEE Press, 2011, 3: 1189-1193. [13] SHRIVAKSHAN G T, CHANDRASEKAR C.A comparison of various edge detection techniques used in image processing[J].International Journal of Computer Science Issues, 2012, 9(5):268-276. http://d.old.wanfangdata.com.cn/OAPaper/oai_doaj-articles_f9703695e3ca9606a91035fec902cf1b [14] YANG Y, PARK D S, HUANG S, et al.Fusion of MT and MR images using an improved wavelet based method[J].Journal of X-ray Science and Technology, 2010, 18(2):157-170. http://www.ncbi.nlm.nih.gov/pubmed/20495243 [15] HUANG W, JING Z.Evaluation of focus measures in multi-focus image fusion[J].Pattern Recognition Letters, 2007, 28(4):493-500. doi: 10.1016/j.patrec.2006.09.005