HEGMP: IMPROVING SIBI HAND GESTURE IMAGE QUALITY COMPARED TO AHE AND CLAHE USING CNN

Abstract

Image visibility with low lighting can cause loss of visual detail in the image, which can obscure information. The research aims to increase the visibility of hand gesture images by contributing to increasing the accuracy of CNN recognition methods. To obtain the best method, the performance of the HE-GMP evaluation method (Histogram Equalization, Gaussian Filter, Median Filter, and Prewitt Edge Detection) is compared with the AHE and CLAHE methods.  The HE-GMP method produces more effective image visibility to improve the quality of SIBI hand gesture information, clarity, sharpness, contrast, brightness. provides detail preservation in areas with high contrast and reduces noise in homogeneous areas, the high increase in color saturation means that hand gesture objects displaying the letters according to SIBI can be recognized easily when compared to the AHE-GMP and CLAHE-GMP methods. The improvement in visual quality of the HE-GMP method with MSE, PSNR, AMBE and Entropy measurement metrics is more significant than the AHE-GMP and CLAHE-GMP methods. Increasing the visibility of hand gesture images with low lighting involves 5 classes, each class consisting of 520 visual data, The HE-GMP method has contributed to increasing accuracy by 62.89% which is more significant compared to the AHE-GMP method of 12.48%, and CLAHE-GMP 20.54%. This research contributes to increasing the accuracy of the CNN method for visual data input in extreme lighting.