基于分块阈值和边缘检测的叶片分割算法

Leaves Segmentation Algorithm Based on Block Threshold and Edge Detection

针对复杂的植物叶片图像,提出了一种分块阈值、边缘检测相结合的图像分割算法。首先,根据预先确定的子块的大小,把整幅图像划分成若干数目的子块,对每个子块用大津法进行分割,把分割好后得到的子图像拼接起来形成目标图像;然后,用改进的Sobel边缘算子对原图像进行边缘提取分割;最后,把分块阈值得到的结果与边缘检测得到的结果结合起来得到较优的结果;在此基础上再进行腐蚀、填洞等形态学操作,得到最终的分割结果。实验表明:与传统的分块阈值、边缘检测相比较,此算法的抗噪性较好,细节上分割得也较为清楚,具有较好的分割效果。

For complex plant leaf images,a segmentation algorithm was presented by combination of sub-block threshold and edge detection.First,according to a pre-determined sub-block size,the entire image was divided into a certain number of sub-blocks and each sub-block was segmented using the Otsu method,the sub-image after segmentation spliced together to form a target image.Then,the original image edges were extracted by improved sobel operator.The two above segmentation results of original image were combined together for better results,finally morphological operations of corrosion and fill holes were applied for the final segmentation result.Experimental results show that comparing with traditional block threshold,edge detection our algorithm not only has better characteristic of noise resistance and show more clearly details of segmentation results,but also have a better segmentation results.