Rotation-invariant and robust multiple-2D-object detection using steerable pyramid denoising and optimized circular harmonic filters

We address the problem of automated detection of biological macromolecules imaged with an electron microscope. These types of images are characterized by a low signal-to-noise ratio (0 dB), a low and varying contrast, in-plane rotated and densely distributed objects in the scene, strong object deformations, and the possible occurrence of false class objects and object agglomerations. We propose a two-step algorithm for rotation invariant and robust multiple 2D-object detection. First, we denoise the image by shrinking the coefficients of a steerable pyramid transform. Second, we employ a correlation-based object detection consisting of a constrained circular harmonic filter. Experiments on our imagery show that this denoising procedure prior to the detection significantly reduces the number of erroneous findings, thus increasing the reliability of object detectors. We elucidate a method to quantify the loss of information due to denoising artifacts on data with an unknown signal-noise relationship. Our experiments prove steerable pyramid denoising to perform well in reconstructing signal out of noisy data