CAJ | 학술논문

为克服传统 Retinex算法没有解决的噪声问题，提出了一种基于 Retinex和 BM3D的图像增强算法。该方法将 BM3D去噪环节加入到传统 Retinex算法中。首先，计算出图像各像素之间的相对明暗关系，进而对待处理图像中的每个像素点灰度值进行校正；然后，通过在相邻图像块中搜索相似块，组成一个三维矩阵，并在三维空间进行滤波处理，得到块预估计值；最后，对图像中每个点进行加权得到最终的处理结果。实验结果表明，该算法既能显示阴影、光照区域中的细节，又能有效消除图像中含有的噪声，克服了传统 Retinex图像增强算法的不足。
위극복전통 Retinex산법몰유해결적조성문제，제출료일충기우 Retinex화 BM3D적도상증강산법。해방법장 BM3D거조배절가입도전통 Retinex산법중。수선，계산출도상각상소지간적상대명암관계，진이대대처리도상중적매개상소점회도치진행교정；연후，통과재상린도상괴중수색상사괴，조성일개삼유구진，병재삼유공간진행려파처리，득도괴예고계치；최후，대도상중매개점진행가권득도최종적처리결과。실험결과표명，해산법기능현시음영、광조구역중적세절，우능유효소제도상중함유적조성，극복료전통 Retinex도상증강산법적불족。
In order to overcome the noise problem of traditional Retinex algorithms, we propose an image enhancement algorithm based on Retinex theory and Blocking Matching 3D (BM3D). This new algorithm adds the denoising links of BM3D into the traditional Retinex algorithm. First, it calculates the relationship of relative lightness between each pixel and corrects the value of each pixel of the image to be handled. Then, through searching similar block among the adjacent tiles, it forms a three-dimensional matrix and filters in 3-D space and gets a preliminarily estimated value. At last, it gets the final result by the weight of each pixel. The experimental results show that the proposed method not only displays the details of the shadow area and lighting areas, but also eliminates the noise contained in the image, overcoming the drawback of traditional enhancement algorithms of Retinex.