Flood fill mean shift: A robust segmentation algorithm

Abstract

In this paper, the flood fill mean shift (FFMS) is introduced. This algorithm is developed for robust segmentation by improving the mean shift (MS) through the flood fill (FF) technique, instead of relying on spatial bandwidth. Due to this exchange, the FFMS involves only one parameter, the range bandwidth, which is not sensitive and is able to acquire global characteristics. If the image parts affected by the illumination changes are sufficiently small and their boundaries are not clear, the illumination effects do not influence the mode seeking procedure of the proposed FFMS. To prove the usefulness and the validity of our algorithm, we present several experiments and analysis of the results.

Application of independent components analysis with the JADE algorithm and NIR hyperspectral imaging for revealing food adulteration

“…Once the images with enhanced contrast are obtained for different components present in a mixed image, different feature extraction algorithms can be applied to locate the exact area of the interesting pixels within the image. In the present work, the connected components labelling algorithm (Di Stefano and Bulgarelli, 1999) followed by the flood fill method to extract object contours (Lee and Kang, 2010), from the image processing toolbox of MATLAB (R2014a), was used to isolate the pixels of peanut particles from the background in the score images. The threshold used to separate the peanut particles from the background was 0.10 and was set by successive trials until the optimal value corresponding to the perfect detection of the overall peanut particles.…”

“…The output BW image replaces all pixels in the input image with luminance greater than threshold level (0.10) with the value 1 (white) and replaces all other pixels with the value 0 (black). The next step consisted in image segmentation with a connected components labelling algorithm (Di Stefano and Bulgarelli, 1999) followed by a flood fill method (Lee and Kang, 2010) to extract object contours. The regionprops function of Matlab Image Processing Toolbox measures different set of properties for each connected component and allows one to retrieve the centroid coordinates of peanut-dots detected in the image.…”

Augmented World: Real Time Gesture Based Image Processing Tool with Intel Real Sense™ Technology

“…Past works on Hand gesture based UI control [31,32] rely on mapping the 2D hand coordinates to mouse position and synthesizing the "click" operation through certain gestures like a fist. The major problem with such a system is that users can use only one hand to control the cursor.…”

A hybrid technique for overlapped chromosome segmentation of G-band mataspread images automatic

“…In this paper used the chromosome images edge detection algorithm (CEDA) [8]. It's used the image enhancement methods in to the edge detection method such as Flood-Fill [12], Erosion [13], and Canny method [14]. The CEDA realized the accuracy detection of chromosome edge, and effectively suppress the effect of noise, and to ensure the continuity, integrity, and location accuracy.…”