Print ISSN: 1681-6900

Online ISSN: 2412-0758

Keywords : SUSAN

Word Retrieval based on FREAK Descriptor to Identify the Image of the English Letter that Corresponds to the First Letter of the Word

Ekhlas F. Naser

Engineering and Technology Journal, 2020, Volume 38, Issue 3B, Pages 150-160
DOI: 10.30684/etj.v38i3B.1511

For the reason of colossal technological developments, the requirement of image information methods became a significant issue. The aim of this research was to retrieve the word based on Fast Retina Key-points (FREAK) descriptor .The suggested system consists of four stages. In the first stage, the images of English letters are loaded. Points are detected via SUSAN in the second stage. FREAK used in the third stage and then a database was created containing 26 English letters. The image to be tested was entered and the points are extracted in the fourth stage and then Manhattan distance was used to calculate the distance between the value of the test image descriptors and all the values of the descriptors in a database. The experimental results show that the precision and the recall values were high for retrieval of the words when using SUSAN because it extracts a large number of interest points compared to the Harris method. For example, for the letter H was 104 with SUSAN while it was 42 for Harris, therefore; the precision for retrieval of the word Hour was 89% and recall was 93% when using SUSAN while precision was 77% and recall was 80% when using Harris.

Improvement of Corner Detection Algorithms (Harris, FAST and SUSAN) Based on Reduction of Features Space and Complexity Time

A.A. Karim; E. F. Nasser

Engineering and Technology Journal, 2017, Volume 35, Issue 2, Pages 112-118

The active detection for gratifying features can be a definitive pace for computer vision in different tasks. Corners become more preferable models because of their two dimensional constrain; two dimensional limitations and algorithms can be rapid to detect them. Corners in images form significant information. Elicitation corners precisely are significant for processing image data to minimize a lot of computations. This paper can be used three vastly algorithms for detection the corner in images improvement Harris, improvement FAST, and improvement SUSAN which are based on two criteria for comparison to minimize the space of interest features and runtime reduction. From that, it can conclude that the algorithm of improvement FAST was outstanding to improvement Harris and improvement SUSAN algorithms on these criteria. FAST, SUSAN and Harris algorithms for corner detected were improved by applying Haar transform and choosing an adaptive gray difference threshold. Improvement FAST, has been offered which can be exceeded the previous two algorithms, improvement Harris and improvement SUSAN in both less run time and small features space. For example, the time taken by car image is 0.0005 second to extract the features using improvement FAST algorithm, which is much less than that used by the SUSAN and Harris algorithms. Improvement Harris takes 0.0074second and SUSAN takes 0.0096 second.