利用粗糙集和属性直方图的图像增强方法

利用粗糙集理论进行图像增强,子图的划分是关键。属性直方图是对直方图概念的推广,是一种由先验知识约束的直方图;将它用于子图的划分,在此基础上提出了一种基于粗糙集理论和属性直方图的图像增强方法。该方法利用属性直方图的 Otsu 算法确定灰度阈值,根据灰度阈值利用不可分辨关系,将图像划分为背景子图、目标子图和噪声子图,对去噪后背景子图和目标子图进行增强变换,并将它们合并得到增强图像。将该方法用于一种海底小目标图像增强。实验结果表明该方法处理增益为 11dB,明显地增强了图像,且不损害图像的边缘。该方法适用于图像有某种先验知识的场合。

Image enhancement using rough sets and bound histogram

It is a crucial problem to partition an image into different sub-images when rough sets theory is applied in image enhancement. The improved histogram, a histogram bound by some prior knowledge, is used for partitioning an image into different sub-images. Furthermore, an image enhancement method based on rough sets and bound histogram is proposed. There are three steps in the method. First, the gray-level threshold is determined by Otsu algorithm based on bound histogram. Second, based on the indiscernible relation, according to the threshold, an image is partitioned into sub-images for background, object and noise. Third, the denoised sub-images of background and object are enhanced respectively and they are combined to form a final enhanced image. The proposed method was applied to the image of a small underwater target. Experimental results show that the SNR (signal-to-noise ratio) increases by 11dB and the image is remarkably enhanced and the boundaries of a region of interest keep unchanged in shape. The method is applicable to images with some prior knowledge.