Application of Wavelet Transform in Vesicle Detection from Biological Fluorescence Images
应用小波变换检测细胞显微荧光图像中的囊泡

This paper present a wavelet transform method to detect vesicles in fluorescence images from biological experiments. It based on à trous wavelet transform to detect vesicles in fluorescence images. To get different wavelet planes, three levels wavelet transform using original image was made, and then median absolute deviation estimate σ was calculated for each wavelet plane. With this σ, a hard threshold method (t=kσ/0.67) was applied to filter out the noise coefficients of each wavelet plane, which allowed to enhance multiscale peaks due to spots. Vesicles were detected via setting a threshold in reconstructed image combining information from different levels of wavelet planes. Meanwhile, the authors compared the "rolling ball" algorithm with the wavelet transform algorithm in their effectiveness of vesicle detection, and found that wavelet transform algorithm is more sensitive in low signal to noise ratio images, more stable for different sizes signal, and better fidelity about the details of signal.