| Summary: | Natural Feature Tracking (NFT) in Augmented Reality (AR) applications use feature detection and a feature matching approach to aligning virtual objects in a real environment. Thus, this tracking detects and compares features that are naturally found in the image (query of image) with the visible feature in the real environment. Therefore, the query of an image must contain good features to track. One of the representing natural features that is easily found in the image is in the corner, and a feature from Accelerated Segment Test (FAST) is one of the fastest corner detectors. However, the FAST corner uses the intensity of the grayscale pixel to determine the candidate corner. Hence, the intensity greatly affects the detection result. Therefore, FAST corner uses the grayscale conversion process to changes the color image into a grayscale image. Thus, the conversion process can lose some details of the images, such as sharpness, shadow, and color image structure. Hence, this process will affect the result of FAST corner to find the feature corner. Besides, Contrast Enhancement also can improve the quality of low contrast grayscale image. In this paper, there are three techniques of the Contrast Enhancement (CE) method were compared, which are Histogram Equalization (HE), Contrast Limited Adaptive Histogram Equalization (CLAHE), and Colormap. As a result, Colormap is better than HE and CLAHE to extract conner and others feature accurately.
|
|---|