A novel edge detection method with application to the fat content prediction in marbled meat

Spectrogram Enhancement By Edge Detection Approach Applied To Bioacoustics Calls Classification

“…The dynamic power range of the spectrogram is limited to avoid the problem of low level portions of the spectrogram expanding and thereby obscuring the detail of the energetic portions. Afterwards, the crest factor image is calculated as a smoothed version of the original spectrogram image, hence escaping the application of smoothing Gaussian filters and their drawbacks [26].Based on the edge detection algorithm presented in [28], the sound patterns in the crest factor image are detected. Afterwards, the original power values of the patterns edges and their interior are reconstructed, while the power values of the patterns surrounding are eliminated, as they represent the attached noises whether attached to the sound or generated during the frequency domain transformation.…”

“…As a result, the crest factor image is obtained, by gathering the local crest factors calculated during the sliding of the mask. Although this edge detection algorithm follows the one presented in [28], it is applied to the crest factor image instead of combination of energy and skewness images, because this combination presents both strong edges (output of the energy feature) and weak edges (output of the skewness feature). Hence, the noises are also detected as patterns, displayed in figure (5a)with a signal to noise ratio (SNR) -given in equation (3) -of1.95.…”

“…Two features from the limited dynamic range spectrogram of figure (3), (a) combination of energy and skewness images, following the algorithm presented in [28], and (b) the crest factor image.…”

“…Finally, the edges image is formed by the values of edges strength (E), and executed by the nonmaximum suppression algorithm and flux equilibrium check [28], to suppress thick edges to one pixel width and fill the missing pixels in the edge direction. Consequently, the final edges image, given in figure (6) is produced, which separates the patterns from surrounding noise.…”

Image processing based method for characterization of the fat/meat ratio and fat distribution of pork and beef samples

“…Jackman and Sun (2009) combined thresholding and clustering to improve segmentation. In some publications, the marbling structure was analysed as well (Taraichi et al, 2002, Hussein, 2011). Barbin and Sun (2011) call the attention on the advantages of extending the wavelength range of the experiments to NIR.…”