Performance Evaluation of CLAHE-Enhanced Edge Detection on Low-Light Faces

Abstract

Edge detection is an important early stage in an image processing-based face detection system. However, the quality of edge detection is highly dependent on the lighting and contrast of the input image. A common problem is the low contrast quality of facial images, which causes edge detection results to be suboptimal, especially in low-light images. This study evaluates the effect of the use of the Contrast Limited Adaptive Histogram Equalization (CLAHE) method on edge detection performance using Canny operators. Two scenarios were tested: edge detection without preprocessing and edge detection after image processing with CLAHE. Evaluation was carried out using two metrics: the number of contours and the total area of the contours of the detected results. The test results showed that the use of CLAHE consistently increased the number of contours and stabilized the contour area distribution, indicating an increased sensitivity to facial edge details. Although an increase in the number of contours can increase the risk of noise detection, the results suggest that CLAHE is able to clarify facial structures that were previously uncaptured. CLAHE has proven to be effective as an image enhancement method in edge detection-based facial detection systems