Defect detection based on extreme edge of defective region histogram

Automatic thresholding has been used by many applications in image processing and pattern recognition systems. Specific attention was given during inspection for quality control purposes in various industries like steel processing and textile manufacturing. Automatic thresholding problem has been addressed well by the commonly used Otsu method, which provides suitable results for thresholding images based on a histogram of bimodal distribution. However, the Otsu method fails when the histogram is unimodal or close to unimodal. Defects have different shapes and sizes, ranging from very small to large. The gray-level distributions of the image histogram can vary between unimodal and multimodal. Furthermore, Otsu-revised methods, like the valley-emphasis method and the background histogram mode extents, which overcome the drawbacks of the Otsu method, require preprocessing steps and fail to use the general threshold for multimodal defects. This study proposes a new automatic thresholding algorithm based on the acquisition of the defective region histogram and the selection of its extreme edge as the threshold value to segment all defective objects in the foreground from the image background. To evaluate the proposed defect-detection method, common standard images for experimentation were used. Experimental results of the proposed method show that the proposed method outperforms the current methods in terms of defect detection.