Performance Evaluation of Blob Detection Techniques Using Image Processing

Blob detection is a primary requirement in computer vision and image processing tasks. Unique visual traits are obtained by identifying blobs in an image. Variations in colour, texture, intensity, or shape are just examples of these attributes, which make blob detection a flexible tool for a variety of image analysis applications. The technique of separating the objects in the binary image is called Binary Large Object (blob) extraction. Every representation technique provides distinct insights into the properties of the blob, which facilitates further processing and interpretation. Techniques for classifying blobs according to their size, shape, texture, and colour are presented, highlighting their importance involving object recognition and classification. The categorization of blob detection techniques presented in this work includes, encompassing Laplacian of Gaussian (LoG), Difference of Gaussian (DoG), Difference of Hessian (DOH), and Connected Components (Labelling). The results demonstrate that LoG and DoG detect a higher number of blobs, with LoG detecting the most (1,366 blobs), while DoH focuses on larger blobs but with a smaller count (777 blobs). Moreover, DoH had the largest mean blob size and higher variability in blob size, while DoG exhibited the fastest detection time (4.27 seconds).